src/sys_fs_cgroup.c

   1 #include "common.h"
   2
   3 // ----------------------------------------------------------------------------
   4 // cgroup globals
   5
   6 static int cgroup_enable_cpuacct_stat = CONFIG_ONDEMAND_ONDEMAND;
   7 static int cgroup_enable_cpuacct_usage = CONFIG_ONDEMAND_ONDEMAND;
   8 static int cgroup_enable_memory = CONFIG_ONDEMAND_ONDEMAND;
   9 static int cgroup_enable_devices = CONFIG_ONDEMAND_ONDEMAND;
  10 static int cgroup_enable_blkio = CONFIG_ONDEMAND_ONDEMAND;
  11 static int cgroup_enable_systemd_services = CONFIG_ONDEMAND_YES;
  12 static int cgroup_enable_new_cgroups_detected_at_runtime = 1;
  13 static int cgroup_check_for_new_every = 10;
  14 static int cgroup_update_every = 1;
  15
  16 static char *cgroup_cpuacct_base = NULL;
  17 static char *cgroup_blkio_base = NULL;
  18 static char *cgroup_memory_base = NULL;
  19 static char *cgroup_devices_base = NULL;
  20
  21 static int cgroup_root_count = 0;
  22 static int cgroup_root_max = 500;
  23 static int cgroup_max_depth = 0;
  24
  25 static SIMPLE_PATTERN *disabled_cgroups_patterns = NULL;
  26 static SIMPLE_PATTERN *disabled_cgroup_paths = NULL;
  27 static SIMPLE_PATTERN *disabled_cgroup_renames = NULL;
  28 static SIMPLE_PATTERN *systemd_services_cgroups = NULL;
  29
  30 static char *cgroups_rename_script = PLUGINS_DIR "/cgroup-name.sh";
  31
  32 static uint32_t Read_hash = 0;
  33 static uint32_t Write_hash = 0;
  34 static uint32_t Sync_hash = 0;
  35 static uint32_t Async_hash = 0;
  36 static uint32_t Total_hash = 0;
  37 static uint32_t user_hash = 0;
  38 static uint32_t system_hash = 0;
  39 static uint32_t cache_hash = 0;
  40 static uint32_t rss_hash = 0;
  41 static uint32_t rss_huge_hash = 0;
  42 static uint32_t mapped_file_hash = 0;
  43 static uint32_t writeback_hash = 0;
  44 static uint32_t dirty_hash = 0;
  45 static uint32_t swap_hash = 0;
  46 static uint32_t pgpgin_hash = 0;
  47 static uint32_t pgpgout_hash = 0;
  48 static uint32_t pgfault_hash = 0;
  49 static uint32_t pgmajfault_hash = 0;
  50 static uint32_t inactive_anon_hash = 0;
  51 static uint32_t active_anon_hash = 0;
  52 static uint32_t inactive_file_hash = 0;
  53 static uint32_t active_file_hash = 0;
  54 static uint32_t unevictable_hash = 0;
  55 static uint32_t hierarchical_memory_limit_hash = 0;
  56 static uint32_t total_cache_hash = 0;
  57 static uint32_t total_rss_hash = 0;
  58 static uint32_t total_rss_huge_hash = 0;
  59 static uint32_t total_mapped_file_hash = 0;
  60 static uint32_t total_writeback_hash = 0;
  61 static uint32_t total_dirty_hash = 0;
  62 static uint32_t total_swap_hash = 0;
  63 static uint32_t total_pgpgin_hash = 0;
  64 static uint32_t total_pgpgout_hash = 0;
  65 static uint32_t total_pgfault_hash = 0;
  66 static uint32_t total_pgmajfault_hash = 0;
  67 static uint32_t total_inactive_anon_hash = 0;
  68 static uint32_t total_active_anon_hash = 0;
  69 static uint32_t total_inactive_file_hash = 0;
  70 static uint32_t total_active_file_hash = 0;
  71 static uint32_t total_unevictable_hash = 0;
  72
  73 void read_cgroup_plugin_configuration() {
  74     Read_hash = simple_hash("Read");
  75     Write_hash = simple_hash("Write");
  76     Sync_hash = simple_hash("Sync");
  77     Async_hash = simple_hash("Async");
  78     Total_hash = simple_hash("Total");
  79     user_hash = simple_hash("user");
  80     system_hash = simple_hash("system");
  81     cache_hash = simple_hash("cache");
  82     rss_hash = simple_hash("rss");
  83     rss_huge_hash = simple_hash("rss_huge");
  84     mapped_file_hash = simple_hash("mapped_file");
  85     writeback_hash = simple_hash("writeback");
  86     dirty_hash = simple_hash("dirty");
  87     swap_hash = simple_hash("swap");
  88     pgpgin_hash = simple_hash("pgpgin");
  89     pgpgout_hash = simple_hash("pgpgout");
  90     pgfault_hash = simple_hash("pgfault");
  91     pgmajfault_hash = simple_hash("pgmajfault");
  92     inactive_anon_hash = simple_hash("inactive_anon");
  93     active_anon_hash = simple_hash("active_anon");
  94     inactive_file_hash = simple_hash("inactive_file");
  95     active_file_hash = simple_hash("active_file");
  96     unevictable_hash = simple_hash("unevictable");
  97     hierarchical_memory_limit_hash = simple_hash("hierarchical_memory_limit");
  98     total_cache_hash = simple_hash("total_cache");
  99     total_rss_hash = simple_hash("total_rss");
 100     total_rss_huge_hash = simple_hash("total_rss_huge");
 101     total_mapped_file_hash = simple_hash("total_mapped_file");
 102     total_writeback_hash = simple_hash("total_writeback");
 103     total_dirty_hash = simple_hash("total_dirty");
 104     total_swap_hash = simple_hash("total_swap");
 105     total_pgpgin_hash = simple_hash("total_pgpgin");
 106     total_pgpgout_hash = simple_hash("total_pgpgout");
 107     total_pgfault_hash = simple_hash("total_pgfault");
 108     total_pgmajfault_hash = simple_hash("total_pgmajfault");
 109     total_inactive_anon_hash = simple_hash("total_inactive_anon");
 110     total_active_anon_hash = simple_hash("total_active_anon");
 111     total_inactive_file_hash = simple_hash("total_inactive_file");
 112     total_active_file_hash = simple_hash("total_active_file");
 113     total_unevictable_hash = simple_hash("total_unevictable");
 114
 115     cgroup_update_every = (int)config_get_number("plugin:cgroups", "update every", rrd_update_every);
 116     if(cgroup_update_every < rrd_update_every)
 117         cgroup_update_every = rrd_update_every;
 118
 119     cgroup_check_for_new_every = (int)config_get_number("plugin:cgroups", "check for new cgroups every", cgroup_check_for_new_every * cgroup_update_every);
 120     if(cgroup_check_for_new_every < cgroup_update_every)
 121         cgroup_check_for_new_every = cgroup_update_every;
 122
 123     cgroup_enable_cpuacct_stat = config_get_boolean_ondemand("plugin:cgroups", "enable cpuacct stat", cgroup_enable_cpuacct_stat);
 124     cgroup_enable_cpuacct_usage = config_get_boolean_ondemand("plugin:cgroups", "enable cpuacct usage", cgroup_enable_cpuacct_usage);
 125     cgroup_enable_memory = config_get_boolean_ondemand("plugin:cgroups", "enable memory", cgroup_enable_memory);
 126     cgroup_enable_blkio = config_get_boolean_ondemand("plugin:cgroups", "enable blkio", cgroup_enable_blkio);
 127
 128     cgroup_enable_systemd_services = config_get_boolean_ondemand("plugin:cgroups", "enable systemd services", cgroup_enable_systemd_services);
 129
 130     char filename[FILENAME_MAX + 1], *s;
 131     struct mountinfo *mi, *root = mountinfo_read(0);
 132
 133     mi = mountinfo_find_by_filesystem_super_option(root, "cgroup", "cpuacct");
 134     if(!mi) mi = mountinfo_find_by_filesystem_mount_source(root, "cgroup", "cpuacct");
 135     if(!mi) {
 136         error("Cannot find cgroup cpuacct mountinfo. Assuming default: /sys/fs/cgroup/cpuacct");
 137         s = "/sys/fs/cgroup/cpuacct";
 138     }
 139     else s = mi->mount_point;
 140     snprintfz(filename, FILENAME_MAX, "%s%s", global_host_prefix, s);
 141     cgroup_cpuacct_base = config_get("plugin:cgroups", "path to /sys/fs/cgroup/cpuacct", filename);
 142
 143     mi = mountinfo_find_by_filesystem_super_option(root, "cgroup", "blkio");
 144     if(!mi) mi = mountinfo_find_by_filesystem_mount_source(root, "cgroup", "blkio");
 145     if(!mi) {
 146         error("Cannot find cgroup blkio mountinfo. Assuming default: /sys/fs/cgroup/blkio");
 147         s = "/sys/fs/cgroup/blkio";
 148     }
 149     else s = mi->mount_point;
 150     snprintfz(filename, FILENAME_MAX, "%s%s", global_host_prefix, s);
 151     cgroup_blkio_base = config_get("plugin:cgroups", "path to /sys/fs/cgroup/blkio", filename);
 152
 153     mi = mountinfo_find_by_filesystem_super_option(root, "cgroup", "memory");
 154     if(!mi) mi = mountinfo_find_by_filesystem_mount_source(root, "cgroup", "memory");
 155     if(!mi) {
 156         error("Cannot find cgroup memory mountinfo. Assuming default: /sys/fs/cgroup/memory");
 157         s = "/sys/fs/cgroup/memory";
 158     }
 159     else s = mi->mount_point;
 160     snprintfz(filename, FILENAME_MAX, "%s%s", global_host_prefix, s);
 161     cgroup_memory_base = config_get("plugin:cgroups", "path to /sys/fs/cgroup/memory", filename);
 162
 163     mi = mountinfo_find_by_filesystem_super_option(root, "cgroup", "devices");
 164     if(!mi) mi = mountinfo_find_by_filesystem_mount_source(root, "cgroup", "devices");
 165     if(!mi) {
 166         error("Cannot find cgroup devices mountinfo. Assuming default: /sys/fs/cgroup/devices");
 167         s = "/sys/fs/cgroup/devices";
 168     }
 169     else s = mi->mount_point;
 170     snprintfz(filename, FILENAME_MAX, "%s%s", global_host_prefix, s);
 171     cgroup_devices_base = config_get("plugin:cgroups", "path to /sys/fs/cgroup/devices", filename);
 172
 173     cgroup_root_max = (int)config_get_number("plugin:cgroups", "max cgroups to allow", cgroup_root_max);
 174     cgroup_max_depth = (int)config_get_number("plugin:cgroups", "max cgroups depth to monitor", cgroup_max_depth);
 175
 176     cgroup_enable_new_cgroups_detected_at_runtime = config_get_boolean("plugin:cgroups", "enable new cgroups detected at run time", cgroup_enable_new_cgroups_detected_at_runtime);
 177
 178     disabled_cgroups_patterns = simple_pattern_create(
 179             config_get("plugin:cgroups", "disable by default cgroups matching", " *.mount "
 180                     " *.partition "
 181                     " *.service "
 182                     " *.slice "
 183                     " *.swap "
 184                     " *.user "
 185                     " / "
 186                     " /docker "
 187                     " /init.scope "
 188                     " /libvirt "
 189                     " /lxc "
 190                     " /lxc/*/ns "                          //  #1397
 191                     " /machine "
 192                     " /machine.slice "
 193                     " /qemu "
 194                     " /system "
 195                     " /system.slice "
 196                     " /systemd "
 197                     " /user "
 198                     " /user.slice "
 199             ), SIMPLE_PATTERN_EXACT);
 200
 201     disabled_cgroup_paths = simple_pattern_create(
 202             config_get("plugin:cgroups", "do not search for cgroups in paths matching"
 203                        , " *-qemu "                             //  #345
 204                             " /init.scope "
 205                             " /system "
 206                             " /systemd "
 207                             " /user "
 208                             " /user.slice "
 209             ), SIMPLE_PATTERN_EXACT);
 210
 211     cgroups_rename_script = config_get("plugin:cgroups", "script to get cgroup names", cgroups_rename_script);
 212
 213     disabled_cgroup_renames = simple_pattern_create(
 214             config_get("plugin:cgroups", "do not run script to rename cgroups matching", " / "
 215                     " *.mount "
 216                     " *.partition "
 217                     " *.scope "
 218                     " *.service "
 219                     " *.slice "
 220                     " *.swap "
 221                     " *.user "
 222             ), SIMPLE_PATTERN_EXACT);
 223
 224     if(cgroup_enable_systemd_services)
 225         systemd_services_cgroups = simple_pattern_create(
 226                 config_get("plugin:cgroups", "cgroups to match as systemd services"
 227                            , " !/system.slice/*/*.service /system.slice/*.service "
 228                 ), SIMPLE_PATTERN_EXACT);
 229
 230     mountinfo_free(root);
 231 }
 232
 233 // ----------------------------------------------------------------------------
 234 // cgroup objects
 235
 236 struct blkio {
 237     int updated;
 238
 239     char *filename;
 240
 241     unsigned long long Read;
 242     unsigned long long Write;
 243 /*
 244     unsigned long long Sync;
 245     unsigned long long Async;
 246     unsigned long long Total;
 247 */
 248 };
 249
 250 // https://www.kernel.org/doc/Documentation/cgroup-v1/memory.txt
 251 struct memory {
 252     int updated;
 253
 254     char *filename;
 255
 256     int has_dirty_swap;
 257
 258     unsigned long long cache;
 259     unsigned long long rss;
 260     unsigned long long rss_huge;
 261     unsigned long long mapped_file;
 262     unsigned long long writeback;
 263     unsigned long long dirty;
 264     unsigned long long swap;
 265     unsigned long long pgpgin;
 266     unsigned long long pgpgout;
 267     unsigned long long pgfault;
 268     unsigned long long pgmajfault;
 269 /*
 270     unsigned long long inactive_anon;
 271     unsigned long long active_anon;
 272     unsigned long long inactive_file;
 273     unsigned long long active_file;
 274     unsigned long long unevictable;
 275     unsigned long long hierarchical_memory_limit;
 276     unsigned long long total_cache;
 277     unsigned long long total_rss;
 278     unsigned long long total_rss_huge;
 279     unsigned long long total_mapped_file;
 280     unsigned long long total_writeback;
 281     unsigned long long total_dirty;
 282     unsigned long long total_swap;
 283     unsigned long long total_pgpgin;
 284     unsigned long long total_pgpgout;
 285     unsigned long long total_pgfault;
 286     unsigned long long total_pgmajfault;
 287     unsigned long long total_inactive_anon;
 288     unsigned long long total_active_anon;
 289     unsigned long long total_inactive_file;
 290     unsigned long long total_active_file;
 291     unsigned long long total_unevictable;
 292 */
 293
 294     int usage_in_bytes_updated;
 295     char *filename_usage_in_bytes;
 296     unsigned long long usage_in_bytes;
 297
 298     int msw_usage_in_bytes_updated;
 299     char *filename_msw_usage_in_bytes;
 300     unsigned long long msw_usage_in_bytes;
 301
 302     int failcnt_updated;
 303     char *filename_failcnt;
 304     unsigned long long failcnt;
 305 };
 306
 307 // https://www.kernel.org/doc/Documentation/cgroup-v1/cpuacct.txt
 308 struct cpuacct_stat {
 309     int updated;
 310
 311     char *filename;
 312
 313     unsigned long long user;
 314     unsigned long long system;
 315 };
 316
 317 // https://www.kernel.org/doc/Documentation/cgroup-v1/cpuacct.txt
 318 struct cpuacct_usage {
 319     int updated;
 320
 321     char *filename;
 322
 323     unsigned int cpus;
 324     unsigned long long *cpu_percpu;
 325 };
 326
 327 #define CGROUP_OPTIONS_DISABLED_DUPLICATE   0x00000001
 328 #define CGROUP_OPTIONS_SYSTEM_SLICE_SERVICE 0x00000002
 329
 330 struct cgroup {
 331     uint32_t options;
 332
 333     char available;      // found in the filesystem
 334     char enabled;        // enabled in the config
 335
 336     char *id;
 337     uint32_t hash;
 338
 339     char *chart_id;
 340     uint32_t hash_chart;
 341
 342     char *chart_title;
 343
 344     struct cpuacct_stat cpuacct_stat;
 345     struct cpuacct_usage cpuacct_usage;
 346
 347     struct memory memory;
 348
 349     struct blkio io_service_bytes;              // bytes
 350     struct blkio io_serviced;                   // operations
 351
 352     struct blkio throttle_io_service_bytes;     // bytes
 353     struct blkio throttle_io_serviced;          // operations
 354
 355     struct blkio io_merged;                     // operations
 356     struct blkio io_queued;                     // operations
 357
 358     // per cgroup charts
 359     RRDSET *st_cpu;
 360     RRDSET *st_cpu_per_core;
 361     RRDSET *st_mem;
 362     RRDSET *st_writeback;
 363     RRDSET *st_mem_activity;
 364     RRDSET *st_pgfaults;
 365     RRDSET *st_mem_usage;
 366     RRDSET *st_mem_failcnt;
 367     RRDSET *st_io;
 368     RRDSET *st_serviced_ops;
 369     RRDSET *st_throttle_io;
 370     RRDSET *st_throttle_serviced_ops;
 371     RRDSET *st_queued_ops;
 372     RRDSET *st_merged_ops;
 373
 374     // services
 375     RRDDIM *rd_cpu;
 376     RRDDIM *rd_mem_usage;
 377
 378     RRDDIM *rd_io_service_bytes_read;
 379     RRDDIM *rd_io_serviced_read;
 380     RRDDIM *rd_throttle_io_read;
 381     RRDDIM *rd_throttle_io_serviced_read;
 382     RRDDIM *rd_io_queued_read;
 383     RRDDIM *rd_io_merged_read;
 384
 385     RRDDIM *rd_io_service_bytes_write;
 386     RRDDIM *rd_io_serviced_write;
 387     RRDDIM *rd_throttle_io_write;
 388     RRDDIM *rd_throttle_io_serviced_write;
 389     RRDDIM *rd_io_queued_write;
 390     RRDDIM *rd_io_merged_write;
 391
 392     struct cgroup *next;
 393
 394 } *cgroup_root = NULL;
 395
 396 // ----------------------------------------------------------------------------
 397 // read values from /sys
 398
 399 static inline void cgroup_read_cpuacct_stat(struct cpuacct_stat *cp) {
 400     static procfile *ff = NULL;
 401
 402     cp->updated = 0;
 403     if(cp->filename) {
 404         ff = procfile_reopen(ff, cp->filename, NULL, PROCFILE_FLAG_DEFAULT);
 405         if(!ff) return;
 406
 407         ff = procfile_readall(ff);
 408         if(!ff) return;
 409
 410         unsigned long i, lines = procfile_lines(ff);
 411
 412         if(lines < 1) {
 413             error("File '%s' should have 1+ lines.", cp->filename);
 414             return;
 415         }
 416
 417         for(i = 0; i < lines ; i++) {
 418             char *s = procfile_lineword(ff, i, 0);
 419             uint32_t hash = simple_hash(s);
 420
 421             if(hash == user_hash && !strcmp(s, "user"))
 422                 cp->user = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 423
 424             else if(hash == system_hash && !strcmp(s, "system"))
 425                 cp->system = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 426         }
 427
 428         cp->updated = 1;
 429
 430         // fprintf(stderr, "READ '%s': user: %llu, system: %llu\n", cp->filename, cp->user, cp->system);
 431     }
 432 }
 433
 434 static inline void cgroup_read_cpuacct_usage(struct cpuacct_usage *ca) {
 435     static procfile *ff = NULL;
 436
 437     ca->updated = 0;
 438     if(ca->filename) {
 439         ff = procfile_reopen(ff, ca->filename, NULL, PROCFILE_FLAG_DEFAULT);
 440         if(!ff) return;
 441
 442         ff = procfile_readall(ff);
 443         if(!ff) return;
 444
 445         if(procfile_lines(ff) < 1) {
 446             error("File '%s' should have 1+ lines but has %u.", ca->filename, procfile_lines(ff));
 447             return;
 448         }
 449
 450         unsigned long i = procfile_linewords(ff, 0);
 451         if(i == 0) return;
 452
 453         // we may have 1 more CPU reported
 454         while(i > 0) {
 455             char *s = procfile_lineword(ff, 0, i - 1);
 456             if(!*s) i--;
 457             else break;
 458         }
 459
 460         if(i != ca->cpus) {
 461             freez(ca->cpu_percpu);
 462             ca->cpu_percpu = mallocz(sizeof(unsigned long long) * i);
 463             ca->cpus = (unsigned int)i;
 464         }
 465
 466         for(i = 0; i < ca->cpus ;i++) {
 467             ca->cpu_percpu[i] = strtoull(procfile_lineword(ff, 0, i), NULL, 10);
 468             // fprintf(stderr, "READ '%s': cpu%d/%d: %llu ('%s')\n", ca->filename, i, ca->cpus, ca->cpu_percpu[i], procfile_lineword(ff, 0, i));
 469         }
 470
 471         ca->updated = 1;
 472     }
 473 }
 474
 475 static inline void cgroup_read_blkio(struct blkio *io) {
 476     static procfile *ff = NULL;
 477
 478     io->updated = 0;
 479     if(io->filename) {
 480         ff = procfile_reopen(ff, io->filename, NULL, PROCFILE_FLAG_DEFAULT);
 481         if(!ff) return;
 482
 483         ff = procfile_readall(ff);
 484         if(!ff) return;
 485
 486         unsigned long i, lines = procfile_lines(ff);
 487
 488         if(lines < 1) {
 489             error("File '%s' should have 1+ lines.", io->filename);
 490             return;
 491         }
 492
 493         io->Read = 0;
 494         io->Write = 0;
 495 /*
 496         io->Sync = 0;
 497         io->Async = 0;
 498         io->Total = 0;
 499 */
 500
 501         for(i = 0; i < lines ; i++) {
 502             char *s = procfile_lineword(ff, i, 1);
 503             uint32_t hash = simple_hash(s);
 504
 505             if(hash == Read_hash && !strcmp(s, "Read"))
 506                 io->Read += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
 507
 508             else if(hash == Write_hash && !strcmp(s, "Write"))
 509                 io->Write += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
 510
 511 /*
 512             else if(hash == Sync_hash && !strcmp(s, "Sync"))
 513                 io->Sync += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
 514
 515             else if(hash == Async_hash && !strcmp(s, "Async"))
 516                 io->Async += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
 517
 518             else if(hash == Total_hash && !strcmp(s, "Total"))
 519                 io->Total += strtoull(procfile_lineword(ff, i, 2), NULL, 10);
 520 */
 521         }
 522
 523         io->updated = 1;
 524         // fprintf(stderr, "READ '%s': Read: %llu, Write: %llu, Sync: %llu, Async: %llu, Total: %llu\n", io->filename, io->Read, io->Write, io->Sync, io->Async, io->Total);
 525     }
 526 }
 527
 528 static inline void cgroup_read_memory(struct memory *mem) {
 529     static procfile *ff = NULL;
 530
 531     mem->updated = 0;
 532     if(mem->filename) {
 533         ff = procfile_reopen(ff, mem->filename, NULL, PROCFILE_FLAG_DEFAULT);
 534         if(!ff) return;
 535
 536         ff = procfile_readall(ff);
 537         if(!ff) return;
 538
 539         unsigned long i, lines = procfile_lines(ff);
 540
 541         if(lines < 1) {
 542             error("File '%s' should have 1+ lines.", mem->filename);
 543             return;
 544         }
 545
 546         for(i = 0; i < lines ; i++) {
 547             char *s = procfile_lineword(ff, i, 0);
 548             uint32_t hash = simple_hash(s);
 549
 550             if(hash == cache_hash && !strcmp(s, "cache"))
 551                 mem->cache = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 552
 553             else if(hash == rss_hash && !strcmp(s, "rss"))
 554                 mem->rss = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 555
 556             else if(hash == rss_huge_hash && !strcmp(s, "rss_huge"))
 557                 mem->rss_huge = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 558
 559             else if(hash == mapped_file_hash && !strcmp(s, "mapped_file"))
 560                 mem->mapped_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 561
 562             else if(hash == writeback_hash && !strcmp(s, "writeback"))
 563                 mem->writeback = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 564
 565             else if(hash == dirty_hash && !strcmp(s, "dirty")) {
 566                 mem->dirty = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 567                 mem->has_dirty_swap = 1;
 568             }
 569
 570             else if(hash == swap_hash && !strcmp(s, "swap")) {
 571                 mem->swap = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 572                 mem->has_dirty_swap = 1;
 573             }
 574
 575             else if(hash == pgpgin_hash && !strcmp(s, "pgpgin"))
 576                 mem->pgpgin = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 577
 578             else if(hash == pgpgout_hash && !strcmp(s, "pgpgout"))
 579                 mem->pgpgout = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 580
 581             else if(hash == pgfault_hash && !strcmp(s, "pgfault"))
 582                 mem->pgfault = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 583
 584             else if(hash == pgmajfault_hash && !strcmp(s, "pgmajfault"))
 585                 mem->pgmajfault = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 586
 587 /*
 588             else if(hash == inactive_anon_hash && !strcmp(s, "inactive_anon"))
 589                 mem->inactive_anon = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 590
 591             else if(hash == active_anon_hash && !strcmp(s, "active_anon"))
 592                 mem->active_anon = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 593
 594             else if(hash == inactive_file_hash && !strcmp(s, "inactive_file"))
 595                 mem->inactive_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 596
 597             else if(hash == active_file_hash && !strcmp(s, "active_file"))
 598                 mem->active_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 599
 600             else if(hash == unevictable_hash && !strcmp(s, "unevictable"))
 601                 mem->unevictable = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 602
 603             else if(hash == hierarchical_memory_limit_hash && !strcmp(s, "hierarchical_memory_limit"))
 604                 mem->hierarchical_memory_limit = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 605
 606             else if(hash == total_cache_hash && !strcmp(s, "total_cache"))
 607                 mem->total_cache = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 608
 609             else if(hash == total_rss_hash && !strcmp(s, "total_rss"))
 610                 mem->total_rss = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 611
 612             else if(hash == total_rss_huge_hash && !strcmp(s, "total_rss_huge"))
 613                 mem->total_rss_huge = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 614
 615             else if(hash == total_mapped_file_hash && !strcmp(s, "total_mapped_file"))
 616                 mem->total_mapped_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 617
 618             else if(hash == total_writeback_hash && !strcmp(s, "total_writeback"))
 619                 mem->total_writeback = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 620
 621             else if(hash == total_dirty_hash && !strcmp(s, "total_dirty"))
 622                 mem->total_dirty = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 623
 624             else if(hash == total_swap_hash && !strcmp(s, "total_swap"))
 625                 mem->total_swap = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 626
 627             else if(hash == total_pgpgin_hash && !strcmp(s, "total_pgpgin"))
 628                 mem->total_pgpgin = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 629
 630             else if(hash == total_pgpgout_hash && !strcmp(s, "total_pgpgout"))
 631                 mem->total_pgpgout = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 632
 633             else if(hash == total_pgfault_hash && !strcmp(s, "total_pgfault"))
 634                 mem->total_pgfault = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 635
 636             else if(hash == total_pgmajfault_hash && !strcmp(s, "total_pgmajfault"))
 637                 mem->total_pgmajfault = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 638
 639             else if(hash == total_inactive_anon_hash && !strcmp(s, "total_inactive_anon"))
 640                 mem->total_inactive_anon = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 641
 642             else if(hash == total_active_anon_hash && !strcmp(s, "total_active_anon"))
 643                 mem->total_active_anon = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 644
 645             else if(hash == total_inactive_file_hash && !strcmp(s, "total_inactive_file"))
 646                 mem->total_inactive_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 647
 648             else if(hash == total_active_file_hash && !strcmp(s, "total_active_file"))
 649                 mem->total_active_file = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 650
 651             else if(hash == total_unevictable_hash && !strcmp(s, "total_unevictable"))
 652                 mem->total_unevictable = strtoull(procfile_lineword(ff, i, 1), NULL, 10);
 653 */
 654         }
 655
 656         // fprintf(stderr, "READ: '%s', cache: %llu, rss: %llu, rss_huge: %llu, mapped_file: %llu, writeback: %llu, dirty: %llu, swap: %llu, pgpgin: %llu, pgpgout: %llu, pgfault: %llu, pgmajfault: %llu, inactive_anon: %llu, active_anon: %llu, inactive_file: %llu, active_file: %llu, unevictable: %llu, hierarchical_memory_limit: %llu, total_cache: %llu, total_rss: %llu, total_rss_huge: %llu, total_mapped_file: %llu, total_writeback: %llu, total_dirty: %llu, total_swap: %llu, total_pgpgin: %llu, total_pgpgout: %llu, total_pgfault: %llu, total_pgmajfault: %llu, total_inactive_anon: %llu, total_active_anon: %llu, total_inactive_file: %llu, total_active_file: %llu, total_unevictable: %llu\n", mem->filename, mem->cache, mem->rss, mem->rss_huge, mem->mapped_file, mem->writeback, mem->dirty, mem->swap, mem->pgpgin, mem->pgpgout, mem->pgfault, mem->pgmajfault, mem->inactive_anon, mem->active_anon, mem->inactive_file, mem->active_file, mem->unevictable, mem->hierarchical_memory_limit, mem->total_cache, mem->total_rss, mem->total_rss_huge, mem->total_mapped_file, mem->total_writeback, mem->total_dirty, mem->total_swap, mem->total_pgpgin, mem->total_pgpgout, mem->total_pgfault, mem->total_pgmajfault, mem->total_inactive_anon, mem->total_active_anon, mem->total_inactive_file, mem->total_active_file, mem->total_unevictable);
 657
 658         mem->updated = 1;
 659     }
 660
 661     mem->usage_in_bytes_updated = 0;
 662     if(mem->filename_usage_in_bytes) {
 663         if(likely(!read_single_number_file(mem->filename_usage_in_bytes, &mem->usage_in_bytes)))
 664             mem->usage_in_bytes_updated = 1;
 665     }
 666
 667     mem->msw_usage_in_bytes_updated = 0;
 668     if(mem->filename_msw_usage_in_bytes) {
 669         if(likely(!read_single_number_file(mem->filename_msw_usage_in_bytes, &mem->msw_usage_in_bytes)))
 670             mem->msw_usage_in_bytes_updated = 1;
 671     }
 672
 673     mem->failcnt_updated = 0;
 674     if(mem->filename_failcnt) {
 675         if(likely(!read_single_number_file(mem->filename_failcnt, &mem->failcnt)))
 676             mem->failcnt_updated = 1;
 677     }
 678 }
 679
 680 static inline void cgroup_read(struct cgroup *cg) {
 681     debug(D_CGROUP, "reading metrics for cgroups '%s'", cg->id);
 682
 683     cgroup_read_cpuacct_stat(&cg->cpuacct_stat);
 684     cgroup_read_cpuacct_usage(&cg->cpuacct_usage);
 685     cgroup_read_memory(&cg->memory);
 686     cgroup_read_blkio(&cg->io_service_bytes);
 687     cgroup_read_blkio(&cg->io_serviced);
 688     cgroup_read_blkio(&cg->throttle_io_service_bytes);
 689     cgroup_read_blkio(&cg->throttle_io_serviced);
 690     cgroup_read_blkio(&cg->io_merged);
 691     cgroup_read_blkio(&cg->io_queued);
 692 }
 693
 694 static inline void read_all_cgroups(struct cgroup *root) {
 695     debug(D_CGROUP, "reading metrics for all cgroups");
 696
 697     struct cgroup *cg;
 698
 699     for(cg = root; cg ; cg = cg->next)
 700         if(cg->enabled && cg->available)
 701             cgroup_read(cg);
 702 }
 703
 704 // ----------------------------------------------------------------------------
 705 // add/remove/find cgroup objects
 706
 707 #define CGROUP_CHARTID_LINE_MAX 1024
 708
 709 static inline char *cgroup_title_strdupz(const char *s) {
 710     if(!s || !*s) s = "/";
 711
 712     if(*s == '/' && s[1] != '\0') s++;
 713
 714     char *r = strdupz(s);
 715     netdata_fix_chart_name(r);
 716
 717     return r;
 718 }
 719
 720 static inline char *cgroup_chart_id_strdupz(const char *s) {
 721     if(!s || !*s) s = "/";
 722
 723     if(*s == '/' && s[1] != '\0') s++;
 724
 725     char *r = strdupz(s);
 726     netdata_fix_chart_id(r);
 727
 728     return r;
 729 }
 730
 731 static inline void cgroup_get_chart_name(struct cgroup *cg) {
 732     debug(D_CGROUP, "looking for the name of cgroup '%s' with chart id '%s' and title '%s'", cg->id, cg->chart_id, cg->chart_title);
 733
 734     pid_t cgroup_pid;
 735     char buffer[CGROUP_CHARTID_LINE_MAX + 1];
 736
 737     snprintfz(buffer, CGROUP_CHARTID_LINE_MAX, "exec %s '%s'", cgroups_rename_script, cg->chart_id);
 738
 739     debug(D_CGROUP, "executing command '%s' for cgroup '%s'", buffer, cg->id);
 740     FILE *fp = mypopen(buffer, &cgroup_pid);
 741     if(fp) {
 742         // debug(D_CGROUP, "reading from command '%s' for cgroup '%s'", buffer, cg->id);
 743         char *s = fgets(buffer, CGROUP_CHARTID_LINE_MAX, fp);
 744         // debug(D_CGROUP, "closing command for cgroup '%s'", cg->id);
 745         mypclose(fp, cgroup_pid);
 746         // debug(D_CGROUP, "closed command for cgroup '%s'", cg->id);
 747
 748         if(s && *s && *s != '\n') {
 749             debug(D_CGROUP, "cgroup '%s' should be renamed to '%s'", cg->id, s);
 750
 751             trim(s);
 752
 753             freez(cg->chart_title);
 754             cg->chart_title = cgroup_title_strdupz(s);
 755
 756             freez(cg->chart_id);
 757             cg->chart_id = cgroup_chart_id_strdupz(s);
 758             cg->hash_chart = simple_hash(cg->chart_id);
 759         }
 760     }
 761     else
 762         error("CGROUP: Cannot popen(\"%s\", \"r\").", buffer);
 763 }
 764
 765 static inline struct cgroup *cgroup_add(const char *id) {
 766     if(!id || !*id) id = "/";
 767     debug(D_CGROUP, "adding to list, cgroup with id '%s'", id);
 768
 769     if(cgroup_root_count >= cgroup_root_max) {
 770         info("Maximum number of cgroups reached (%d). Not adding cgroup '%s'", cgroup_root_count, id);
 771         return NULL;
 772     }
 773
 774     int def = simple_pattern_matches(disabled_cgroups_patterns, id)?0:cgroup_enable_new_cgroups_detected_at_runtime;
 775     struct cgroup *cg = callocz(1, sizeof(struct cgroup));
 776
 777     cg->id = strdupz(id);
 778     cg->hash = simple_hash(cg->id);
 779
 780     cg->chart_title = cgroup_title_strdupz(id);
 781
 782     cg->chart_id = cgroup_chart_id_strdupz(id);
 783     cg->hash_chart = simple_hash(cg->chart_id);
 784
 785     if(!cgroup_root)
 786         cgroup_root = cg;
 787     else {
 788         // append it
 789         struct cgroup *e;
 790         for(e = cgroup_root; e->next ;e = e->next) ;
 791         e->next = cg;
 792     }
 793
 794     cgroup_root_count++;
 795
 796     // fix the chart_id and title by calling the external script
 797     if(!simple_pattern_matches(disabled_cgroup_renames, cg->id) &&
 798        !simple_pattern_matches(disabled_cgroup_renames, cg->chart_id)) {
 799
 800         cgroup_get_chart_name(cg);
 801
 802         debug(D_CGROUP, "cgroup '%s' renamed to '%s' (title: '%s')", cg->id, cg->chart_id, cg->chart_title);
 803     }
 804     else
 805         debug(D_CGROUP, "cgroup '%s' will not be renamed - it matches the list of disabled cgroup renames (will be shown as '%s')", cg->id, cg->chart_id);
 806
 807     int user_configurable = 1;
 808
 809     // check if this cgroup should be a systemd service
 810     if(cgroup_enable_systemd_services) {
 811         if(simple_pattern_matches(systemd_services_cgroups, cg->id) ||
 812                 simple_pattern_matches(systemd_services_cgroups, cg->chart_id)) {
 813             debug(D_CGROUP, "cgroup '%s' with chart id '%s' (title: '%s') matches systemd services cgroups", cg->id, cg->chart_id, cg->chart_title);
 814
 815             char buffer[CGROUP_CHARTID_LINE_MAX + 1];
 816             cg->options |= CGROUP_OPTIONS_SYSTEM_SLICE_SERVICE;
 817
 818             strncpy(buffer, cg->id, CGROUP_CHARTID_LINE_MAX);
 819             char *s = buffer;
 820
 821             //freez(cg->chart_id);
 822             //cg->chart_id = cgroup_chart_id_strdupz(s);
 823             //cg->hash_chart = simple_hash(cg->chart_id);
 824
 825             // skip to the last slash
 826             size_t len = strlen(s);
 827             while(len--) if(unlikely(s[len] == '/')) break;
 828             if(len) s = &s[len + 1];
 829
 830             // remove extension
 831             len = strlen(s);
 832             while(len--) if(unlikely(s[len] == '.')) break;
 833             if(len) s[len] = '\0';
 834
 835             freez(cg->chart_title);
 836             cg->chart_title = cgroup_title_strdupz(s);
 837
 838             cg->enabled = 1;
 839             user_configurable = 0;
 840
 841             debug(D_CGROUP, "cgroup '%s' renamed to '%s' (title: '%s')", cg->id, cg->chart_id, cg->chart_title);
 842         }
 843         else
 844             debug(D_CGROUP, "cgroup '%s' with chart id '%s' (title: '%s') does not match systemd services groups", cg->id, cg->chart_id, cg->chart_title);
 845     }
 846
 847     if(user_configurable) {
 848         // allow the user to enable/disable this individualy
 849         char option[FILENAME_MAX + 1];
 850         snprintfz(option, FILENAME_MAX, "enable cgroup %s", cg->chart_title);
 851         cg->enabled = (char) config_get_boolean("plugin:cgroups", option, def);
 852     }
 853
 854     // detect duplicate cgroups
 855     if(cg->enabled) {
 856         struct cgroup *t;
 857         for (t = cgroup_root; t; t = t->next) {
 858             if (t != cg && t->enabled && t->hash_chart == cg->hash_chart && !strcmp(t->chart_id, cg->chart_id)) {
 859                 if (!strncmp(t->chart_id, "/system.slice/", 14) && !strncmp(cg->chart_id, "/init.scope/system.slice/", 25)) {
 860                     error("Control group with chart id '%s' already exists with id '%s' and is enabled. Swapping them by enabling cgroup with id '%s' and disabling cgroup with id '%s'.",
 861                           cg->chart_id, t->id, cg->id, t->id);
 862                     debug(D_CGROUP, "Control group with chart id '%s' already exists with id '%s' and is enabled. Swapping them by enabling cgroup with id '%s' and disabling cgroup with id '%s'.",
 863                           cg->chart_id, t->id, cg->id, t->id);
 864                     t->enabled = 0;
 865                     t->options |= CGROUP_OPTIONS_DISABLED_DUPLICATE;
 866                 }
 867                 else {
 868                     error("Control group with chart id '%s' already exists with id '%s' and is enabled and available. Disabling cgroup with id '%s'.",
 869                           cg->chart_id, t->id, cg->id);
 870                     debug(D_CGROUP, "Control group with chart id '%s' already exists with id '%s' and is enabled and available. Disabling cgroup with id '%s'.",
 871                           cg->chart_id, t->id, cg->id);
 872                     cg->enabled = 0;
 873                     cg->options |= CGROUP_OPTIONS_DISABLED_DUPLICATE;
 874                 }
 875
 876                 break;
 877             }
 878         }
 879     }
 880
 881     debug(D_CGROUP, "ADDED CGROUP: '%s' with chart id '%s' and title '%s' as %s (default was %s)", cg->id, cg->chart_id, cg->chart_title, (cg->enabled)?"enabled":"disabled", (def)?"enabled":"disabled");
 882
 883     return cg;
 884 }
 885
 886 static inline void cgroup_free(struct cgroup *cg) {
 887     debug(D_CGROUP, "Removing cgroup '%s' with chart id '%s' (was %s and %s)", cg->id, cg->chart_id, (cg->enabled)?"enabled":"disabled", (cg->available)?"available":"not available");
 888
 889     freez(cg->cpuacct_usage.cpu_percpu);
 890
 891     freez(cg->cpuacct_stat.filename);
 892     freez(cg->cpuacct_usage.filename);
 893
 894     freez(cg->memory.filename);
 895     freez(cg->memory.filename_failcnt);
 896     freez(cg->memory.filename_usage_in_bytes);
 897     freez(cg->memory.filename_msw_usage_in_bytes);
 898
 899     freez(cg->io_service_bytes.filename);
 900     freez(cg->io_serviced.filename);
 901
 902     freez(cg->throttle_io_service_bytes.filename);
 903     freez(cg->throttle_io_serviced.filename);
 904
 905     freez(cg->io_merged.filename);
 906     freez(cg->io_queued.filename);
 907
 908     freez(cg->id);
 909     freez(cg->chart_id);
 910     freez(cg->chart_title);
 911
 912     freez(cg);
 913
 914     cgroup_root_count--;
 915 }
 916
 917 // find if a given cgroup exists
 918 static inline struct cgroup *cgroup_find(const char *id) {
 919     debug(D_CGROUP, "searching for cgroup '%s'", id);
 920
 921     uint32_t hash = simple_hash(id);
 922
 923     struct cgroup *cg;
 924     for(cg = cgroup_root; cg ; cg = cg->next) {
 925         if(hash == cg->hash && strcmp(id, cg->id) == 0)
 926             break;
 927     }
 928
 929     debug(D_CGROUP, "cgroup '%s' %s in memory", id, (cg)?"found":"not found");
 930     return cg;
 931 }
 932
 933 // ----------------------------------------------------------------------------
 934 // detect running cgroups
 935
 936 // callback for find_file_in_subdirs()
 937 static inline void found_subdir_in_dir(const char *dir) {
 938     debug(D_CGROUP, "examining cgroup dir '%s'", dir);
 939
 940     struct cgroup *cg = cgroup_find(dir);
 941     if(!cg) {
 942         if(*dir && cgroup_max_depth > 0) {
 943             int depth = 0;
 944             const char *s;
 945
 946             for(s = dir; *s ;s++)
 947                 if(unlikely(*s == '/'))
 948                     depth++;
 949
 950             if(depth > cgroup_max_depth) {
 951                 info("cgroup '%s' is too deep (%d, while max is %d)", dir, depth, cgroup_max_depth);
 952                 return;
 953             }
 954         }
 955         // debug(D_CGROUP, "will add dir '%s' as cgroup", dir);
 956         cg = cgroup_add(dir);
 957     }
 958
 959     if(cg) cg->available = 1;
 960 }
 961
 962 static inline int find_dir_in_subdirs(const char *base, const char *this, void (*callback)(const char *)) {
 963     if(!this) this = base;
 964     debug(D_CGROUP, "searching for directories in '%s' (base '%s')", this?this:"", base);
 965
 966     size_t dirlen = strlen(this), baselen = strlen(base);
 967
 968     int ret = -1;
 969     int enabled = -1;
 970
 971     const char *relative_path = &this[baselen];
 972     if(!*relative_path) relative_path = "/";
 973
 974     DIR *dir = opendir(this);
 975     if(!dir) {
 976         error("Cannot read cgroups directory '%s'", base);
 977         return ret;
 978     }
 979     ret = 1;
 980
 981     callback(relative_path);
 982
 983     struct dirent *de = NULL;
 984     while((de = readdir(dir))) {
 985         if(de->d_type == DT_DIR
 986             && (
 987                 (de->d_name[0] == '.' && de->d_name[1] == '\0')
 988                 || (de->d_name[0] == '.' && de->d_name[1] == '.' && de->d_name[2] == '\0')
 989                 ))
 990             continue;
 991
 992         if(de->d_type == DT_DIR) {
 993             if(enabled == -1) {
 994                 const char *r = relative_path;
 995                 if(*r == '\0') r = "/";
 996
 997                 // do not decent in directories we are not interested
 998                 int def = 1;
 999                 if(simple_pattern_matches(disabled_cgroup_paths, r))
1000                     def = 0;
1001
1002                 // we check for this option here
1003                 // so that the config will not have settings
1004                 // for leaf directories
1005                 char option[FILENAME_MAX + 1];
1006                 snprintfz(option, FILENAME_MAX, "search for cgroups under %s", r);
1007                 option[FILENAME_MAX] = '\0';
1008                 enabled = config_get_boolean("plugin:cgroups", option, def);
1009             }
1010
1011             if(enabled) {
1012                 char *s = mallocz(dirlen + strlen(de->d_name) + 2);
1013                 strcpy(s, this);
1014                 strcat(s, "/");
1015                 strcat(s, de->d_name);
1016                 int ret2 = find_dir_in_subdirs(base, s, callback);
1017                 if(ret2 > 0) ret += ret2;
1018                 freez(s);
1019             }
1020         }
1021     }
1022
1023     closedir(dir);
1024     return ret;
1025 }
1026
1027 static inline void mark_all_cgroups_as_not_available() {
1028     debug(D_CGROUP, "marking all cgroups as not available");
1029
1030     struct cgroup *cg;
1031
1032     // mark all as not available
1033     for(cg = cgroup_root; cg ; cg = cg->next) {
1034         cg->available = 0;
1035     }
1036 }
1037
1038 static inline void cleanup_all_cgroups() {
1039     struct cgroup *cg = cgroup_root, *last = NULL;
1040
1041     for(; cg ;) {
1042         if(!cg->available) {
1043             // enable the first duplicate cgroup
1044             {
1045                 struct cgroup *t;
1046                 for(t = cgroup_root; t ; t = t->next) {
1047                     if(t != cg && t->available && !t->enabled && t->options & CGROUP_OPTIONS_DISABLED_DUPLICATE && t->hash_chart == cg->hash_chart && !strcmp(t->chart_id, cg->chart_id)) {
1048                         debug(D_CGROUP, "Enabling duplicate of cgroup '%s' with id '%s', because the original with id '%s' stopped.", t->chart_id, t->id, cg->id);
1049                         t->enabled = 1;
1050                         t->options &= ~CGROUP_OPTIONS_DISABLED_DUPLICATE;
1051                         break;
1052                     }
1053                 }
1054             }
1055
1056             if(!last)
1057                 cgroup_root = cg->next;
1058             else
1059                 last->next = cg->next;
1060
1061             cgroup_free(cg);
1062
1063             if(!last)
1064                 cg = cgroup_root;
1065             else
1066                 cg = last->next;
1067         }
1068         else {
1069             last = cg;
1070             cg = cg->next;
1071         }
1072     }
1073 }
1074
1075 static inline void find_all_cgroups() {
1076     debug(D_CGROUP, "searching for cgroups");
1077
1078     mark_all_cgroups_as_not_available();
1079
1080     if(cgroup_enable_cpuacct_stat || cgroup_enable_cpuacct_usage) {
1081         if (find_dir_in_subdirs(cgroup_cpuacct_base, NULL, found_subdir_in_dir) == -1) {
1082             cgroup_enable_cpuacct_stat = cgroup_enable_cpuacct_usage = 0;
1083             error("disabled cgroup cpu statistics.");
1084         }
1085     }
1086
1087     if(cgroup_enable_blkio) {
1088         if (find_dir_in_subdirs(cgroup_blkio_base, NULL, found_subdir_in_dir) == -1) {
1089             cgroup_enable_blkio = 0;
1090             error("disabled cgroup blkio statistics.");
1091         }
1092     }
1093
1094     if(cgroup_enable_memory) {
1095         if(find_dir_in_subdirs(cgroup_memory_base, NULL, found_subdir_in_dir) == -1) {
1096             cgroup_enable_memory = 0;
1097             error("disabled cgroup memory statistics.");
1098         }
1099     }
1100
1101     if(cgroup_enable_devices) {
1102         if(find_dir_in_subdirs(cgroup_devices_base, NULL, found_subdir_in_dir) == -1) {
1103             cgroup_enable_devices = 0;
1104             error("disabled cgroup devices statistics.");
1105         }
1106     }
1107
1108     // remove any non-existing cgroups
1109     cleanup_all_cgroups();
1110
1111     struct cgroup *cg;
1112     struct stat buf;
1113     for(cg = cgroup_root; cg ; cg = cg->next) {
1114         // fprintf(stderr, " >>> CGROUP '%s' (%u - %s) with name '%s'\n", cg->id, cg->hash, cg->available?"available":"stopped", cg->name);
1115
1116         if(unlikely(!cg->available))
1117             continue;
1118
1119         debug(D_CGROUP, "checking paths for cgroup '%s'", cg->id);
1120
1121         // check for newly added cgroups
1122         // and update the filenames they read
1123         char filename[FILENAME_MAX + 1];
1124         if(unlikely(cgroup_enable_cpuacct_stat && !cg->cpuacct_stat.filename)) {
1125             snprintfz(filename, FILENAME_MAX, "%s%s/cpuacct.stat", cgroup_cpuacct_base, cg->id);
1126             if(stat(filename, &buf) != -1) {
1127                 cg->cpuacct_stat.filename = strdupz(filename);
1128                 debug(D_CGROUP, "cpuacct.stat filename for cgroup '%s': '%s'", cg->id, cg->cpuacct_stat.filename);
1129             }
1130             else debug(D_CGROUP, "cpuacct.stat file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1131         }
1132
1133         if(unlikely(cgroup_enable_cpuacct_usage && !cg->cpuacct_usage.filename && !(cg->options & CGROUP_OPTIONS_SYSTEM_SLICE_SERVICE))) {
1134             snprintfz(filename, FILENAME_MAX, "%s%s/cpuacct.usage_percpu", cgroup_cpuacct_base, cg->id);
1135             if(stat(filename, &buf) != -1) {
1136                 cg->cpuacct_usage.filename = strdupz(filename);
1137                 debug(D_CGROUP, "cpuacct.usage_percpu filename for cgroup '%s': '%s'", cg->id, cg->cpuacct_usage.filename);
1138             }
1139             else debug(D_CGROUP, "cpuacct.usage_percpu file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1140         }
1141
1142         if(unlikely(cgroup_enable_memory)) {
1143             if(unlikely(!cg->memory.filename) && !(cg->options & CGROUP_OPTIONS_SYSTEM_SLICE_SERVICE)) {
1144                 snprintfz(filename, FILENAME_MAX, "%s%s/memory.stat", cgroup_memory_base, cg->id);
1145                 if(stat(filename, &buf) != -1) {
1146                     cg->memory.filename = strdupz(filename);
1147                     debug(D_CGROUP, "memory.stat filename for cgroup '%s': '%s'", cg->id, cg->memory.filename);
1148                 }
1149                 else
1150                     debug(D_CGROUP, "memory.stat file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1151             }
1152             if(unlikely(!cg->memory.filename_usage_in_bytes)) {
1153                 snprintfz(filename, FILENAME_MAX, "%s%s/memory.usage_in_bytes", cgroup_memory_base, cg->id);
1154                 if(stat(filename, &buf) != -1) {
1155                     cg->memory.filename_usage_in_bytes = strdupz(filename);
1156                     debug(D_CGROUP, "memory.usage_in_bytes filename for cgroup '%s': '%s'", cg->id, cg->memory.filename_usage_in_bytes);
1157                 }
1158                 else
1159                     debug(D_CGROUP, "memory.usage_in_bytes file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1160             }
1161             if(unlikely(!cg->memory.filename_msw_usage_in_bytes)) {
1162                 snprintfz(filename, FILENAME_MAX, "%s%s/memory.msw_usage_in_bytes", cgroup_memory_base, cg->id);
1163                 if(stat(filename, &buf) != -1) {
1164                     cg->memory.filename_msw_usage_in_bytes = strdupz(filename);
1165                     debug(D_CGROUP, "memory.msw_usage_in_bytes filename for cgroup '%s': '%s'", cg->id, cg->memory.filename_msw_usage_in_bytes);
1166                 }
1167                 else
1168                     debug(D_CGROUP, "memory.msw_usage_in_bytes file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1169             }
1170             if(unlikely(!cg->memory.filename_failcnt)) {
1171                 snprintfz(filename, FILENAME_MAX, "%s%s/memory.failcnt", cgroup_memory_base, cg->id);
1172                 if(stat(filename, &buf) != -1) {
1173                     cg->memory.filename_failcnt = strdupz(filename);
1174                     debug(D_CGROUP, "memory.failcnt filename for cgroup '%s': '%s'", cg->id, cg->memory.filename_failcnt);
1175                 }
1176                 else
1177                     debug(D_CGROUP, "memory.failcnt file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1178             }
1179         }
1180
1181         if(unlikely(cgroup_enable_blkio)) {
1182             if(unlikely(!cg->io_service_bytes.filename)) {
1183                 snprintfz(filename, FILENAME_MAX, "%s%s/blkio.io_service_bytes", cgroup_blkio_base, cg->id);
1184                 if(stat(filename, &buf) != -1) {
1185                     cg->io_service_bytes.filename = strdupz(filename);
1186                     debug(D_CGROUP, "io_service_bytes filename for cgroup '%s': '%s'", cg->id, cg->io_service_bytes.filename);
1187                 }
1188                 else debug(D_CGROUP, "io_service_bytes file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1189             }
1190             if(unlikely(!cg->io_serviced.filename)) {
1191                 snprintfz(filename, FILENAME_MAX, "%s%s/blkio.io_serviced", cgroup_blkio_base, cg->id);
1192                 if(stat(filename, &buf) != -1) {
1193                     cg->io_serviced.filename = strdupz(filename);
1194                     debug(D_CGROUP, "io_serviced filename for cgroup '%s': '%s'", cg->id, cg->io_serviced.filename);
1195                 }
1196                 else debug(D_CGROUP, "io_serviced file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1197             }
1198             if(unlikely(!cg->throttle_io_service_bytes.filename)) {
1199                 snprintfz(filename, FILENAME_MAX, "%s%s/blkio.throttle.io_service_bytes", cgroup_blkio_base, cg->id);
1200                 if(stat(filename, &buf) != -1) {
1201                     cg->throttle_io_service_bytes.filename = strdupz(filename);
1202                     debug(D_CGROUP, "throttle_io_service_bytes filename for cgroup '%s': '%s'", cg->id, cg->throttle_io_service_bytes.filename);
1203                 }
1204                 else debug(D_CGROUP, "throttle_io_service_bytes file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1205             }
1206             if(unlikely(!cg->throttle_io_serviced.filename)) {
1207                 snprintfz(filename, FILENAME_MAX, "%s%s/blkio.throttle.io_serviced", cgroup_blkio_base, cg->id);
1208                 if(stat(filename, &buf) != -1) {
1209                     cg->throttle_io_serviced.filename = strdupz(filename);
1210                     debug(D_CGROUP, "throttle_io_serviced filename for cgroup '%s': '%s'", cg->id, cg->throttle_io_serviced.filename);
1211                 }
1212                 else debug(D_CGROUP, "throttle_io_serviced file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1213             }
1214             if(unlikely(!cg->io_merged.filename)) {
1215                 snprintfz(filename, FILENAME_MAX, "%s%s/blkio.io_merged", cgroup_blkio_base, cg->id);
1216                 if(stat(filename, &buf) != -1) {
1217                     cg->io_merged.filename = strdupz(filename);
1218                     debug(D_CGROUP, "io_merged filename for cgroup '%s': '%s'", cg->id, cg->io_merged.filename);
1219                 }
1220                 else debug(D_CGROUP, "io_merged file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1221             }
1222             if(unlikely(!cg->io_queued.filename)) {
1223                 snprintfz(filename, FILENAME_MAX, "%s%s/blkio.io_queued", cgroup_blkio_base, cg->id);
1224                 if(stat(filename, &buf) != -1) {
1225                     cg->io_queued.filename = strdupz(filename);
1226                     debug(D_CGROUP, "io_queued filename for cgroup '%s': '%s'", cg->id, cg->io_queued.filename);
1227                 }
1228                 else debug(D_CGROUP, "io_queued file for cgroup '%s': '%s' does not exist.", cg->id, filename);
1229             }
1230         }
1231     }
1232
1233     debug(D_CGROUP, "done searching for cgroups");
1234     return;
1235 }
1236
1237 // ----------------------------------------------------------------------------
1238 // generate charts
1239
1240 #define CHART_TITLE_MAX 300
1241
1242 void update_services_charts(int update_every,
1243         int do_cpu,
1244         int do_mem_usage,
1245         int do_io,
1246         int do_io_ops,
1247         int do_throttle_io,
1248         int do_throttle_ops /*,
1249         int do_queued_ops,
1250         int do_merged_ops */
1251 ) {
1252     static RRDSET
1253             *st_cpu = NULL,
1254             *st_mem_usage = NULL,
1255
1256             *st_io_read = NULL,
1257             *st_io_serviced_read = NULL,
1258             *st_throttle_io_read = NULL,
1259             *st_throttle_ops_read = NULL,
1260             /*
1261             *st_queued_ops_read = NULL,
1262             *st_merged_ops_read = NULL,
1263             * */
1264
1265             *st_io_write = NULL,
1266             *st_io_serviced_write = NULL,
1267             *st_throttle_io_write = NULL,
1268             *st_throttle_ops_write = NULL/*,
1269             *st_queued_ops_write = NULL,
1270             *st_merged_ops_write = NULL*/;
1271
1272     // create the charts
1273
1274     if(do_cpu) {
1275         if(unlikely(!st_cpu)) {
1276             char title[CHART_TITLE_MAX + 1];
1277
1278             st_cpu = rrdset_find_bytype("services", "cpu");
1279             if(!st_cpu) {
1280                 snprintfz(title, CHART_TITLE_MAX, "Systemd Services CPU utilization (%d%% = %d core%s)", (processors * 100), processors, (processors > 1) ? "s" : "");
1281                 st_cpu = rrdset_create("services", "cpu", NULL, "cpu", "services.cpu", title, "%", 30000, update_every, RRDSET_TYPE_STACKED);
1282             }
1283         }
1284         else rrdset_next(st_cpu);
1285     }
1286
1287     if(do_mem_usage) {
1288         if(unlikely(!st_mem_usage)) {
1289             st_mem_usage = rrdset_find_bytype("services", "mem_usage");
1290             if(!st_mem_usage)
1291                 st_mem_usage = rrdset_create("services", "mem_usage", NULL, "mem", "services.mem_usage", "Systemd Services RAM Usage", "MB", 30001, update_every, RRDSET_TYPE_STACKED);
1292         }
1293         else rrdset_next(st_mem_usage);
1294     }
1295
1296     if(do_io) {
1297         if(unlikely(!st_io_read)) {
1298             st_io_read = rrdset_find_bytype("services", "io_read");
1299             if(!st_io_read)
1300                 st_io_read = rrdset_create("services", "io_read", NULL, "disk", "services.io_read", "Systemd Services Disk Read Bandwidth", "KB/s", 30012, update_every, RRDSET_TYPE_STACKED);
1301         }
1302         else rrdset_next(st_io_read);
1303
1304         if(unlikely(!st_io_write)) {
1305             st_io_write = rrdset_find_bytype("services", "io_write");
1306             if(!st_io_write)
1307                 st_io_write = rrdset_create("services", "io_write", NULL, "disk", "services.io_write", "Systemd Services Disk Write Bandwidth", "KB/s", 30013, update_every, RRDSET_TYPE_STACKED);
1308         }
1309         else rrdset_next(st_io_write);
1310     }
1311
1312     if(do_io_ops) {
1313         if(unlikely(!st_io_serviced_read)) {
1314             st_io_serviced_read = rrdset_find_bytype("services", "io_ops_read");
1315             if(!st_io_serviced_read)
1316                 st_io_serviced_read = rrdset_create("services", "io_ops_read", NULL, "disk", "services.io_ops_read", "Systemd Services Disk Read Operations", "operations/s", 30014, update_every, RRDSET_TYPE_STACKED);
1317         }
1318         else rrdset_next(st_io_serviced_read);
1319
1320         if(unlikely(!st_io_serviced_write)) {
1321             st_io_serviced_write = rrdset_find_bytype("services", "io_ops_write");
1322             if(!st_io_serviced_write)
1323                 st_io_serviced_write = rrdset_create("services", "io_ops_write", NULL, "disk", "services.io_ops_write", "Systemd Services Disk Write Operations", "operations/s", 30015, update_every, RRDSET_TYPE_STACKED);
1324         }
1325         else rrdset_next(st_io_serviced_write);
1326     }
1327
1328     if(do_throttle_io) {
1329         if(unlikely(!st_throttle_io_read)) {
1330             st_throttle_io_read = rrdset_find_bytype("services", "throttle_io_read");
1331             if(!st_throttle_io_read)
1332                 st_throttle_io_read = rrdset_create("services", "throttle_io_read", NULL, "disk", "services.throttle_io_read", "Systemd Services Throttle Disk Read Bandwidth", "KB/s", 30016, update_every, RRDSET_TYPE_STACKED);
1333         }
1334         else rrdset_next(st_throttle_io_read);
1335
1336         if(unlikely(!st_throttle_io_write)) {
1337             st_throttle_io_write = rrdset_find_bytype("services", "throttle_io_write");
1338             if(!st_throttle_io_write)
1339                 st_throttle_io_write = rrdset_create("services", "throttle_io_write", NULL, "disk", "services.throttle_io_write", "Systemd Services Throttle Disk Write Bandwidth", "KB/s", 30017, update_every, RRDSET_TYPE_STACKED);
1340         }
1341         else rrdset_next(st_throttle_io_write);
1342     }
1343
1344     if(do_throttle_ops) {
1345         if(unlikely(!st_throttle_ops_read)) {
1346             st_throttle_ops_read = rrdset_find_bytype("services", "throttle_io_ops_read");
1347             if(!st_throttle_ops_read)
1348                 st_throttle_ops_read = rrdset_create("services", "throttle_io_ops_read", NULL, "disk", "services.throttle_io_ops_read", "Systemd Services Throttle Disk Read Operations", "operations/s", 30018, update_every, RRDSET_TYPE_STACKED);
1349         }
1350         else rrdset_next(st_throttle_ops_read);
1351
1352         if(unlikely(!st_throttle_ops_write)) {
1353             st_throttle_ops_write = rrdset_find_bytype("services", "throttle_io_ops_write");
1354             if(!st_throttle_ops_write)
1355                 st_throttle_ops_write = rrdset_create("services", "throttle_io_ops_write", NULL, "disk", "services.throttle_io_ops_write", "Systemd Services Throttle Disk Write Operations", "operations/s", 30019, update_every, RRDSET_TYPE_STACKED);
1356         }
1357         else rrdset_next(st_throttle_ops_write);
1358     }
1359
1360     // update the values
1361     struct cgroup *cg;
1362     for(cg = cgroup_root; cg ; cg = cg->next) {
1363         if(!cg->available || !cg->enabled || !(cg->options & CGROUP_OPTIONS_SYSTEM_SLICE_SERVICE)) continue;
1364
1365         if(do_cpu && cg->cpuacct_stat.updated) {
1366             if(unlikely(!cg->rd_cpu))
1367                 cg->rd_cpu = rrddim_add(st_cpu, cg->chart_id, cg->chart_title, 100, hz, RRDDIM_INCREMENTAL);
1368
1369             rrddim_set_by_pointer(st_cpu, cg->rd_cpu, cg->cpuacct_stat.user + cg->cpuacct_stat.system);
1370         }
1371
1372         if(do_mem_usage && cg->memory.usage_in_bytes_updated) {
1373             if(unlikely(!cg->rd_mem_usage))
1374                 cg->rd_mem_usage = rrddim_add(st_mem_usage, cg->chart_id, cg->chart_title, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1375
1376             rrddim_set_by_pointer(st_mem_usage, cg->rd_mem_usage, cg->memory.usage_in_bytes);
1377         }
1378
1379         if(do_io && cg->io_service_bytes.updated) {
1380             if(unlikely(!cg->rd_io_service_bytes_read))
1381                 cg->rd_io_service_bytes_read = rrddim_add(st_io_read, cg->chart_id, cg->chart_title, 1, 1024, RRDDIM_INCREMENTAL);
1382
1383             rrddim_set_by_pointer(st_io_read, cg->rd_io_service_bytes_read, cg->io_service_bytes.Read);
1384
1385             if(unlikely(!cg->rd_io_service_bytes_write))
1386                 cg->rd_io_service_bytes_write = rrddim_add(st_io_write, cg->chart_id, cg->chart_title, 1, 1024, RRDDIM_INCREMENTAL);
1387
1388             rrddim_set_by_pointer(st_io_write, cg->rd_io_service_bytes_write, cg->io_service_bytes.Write);
1389         }
1390
1391         if(do_io_ops && cg->io_serviced.updated) {
1392             if(unlikely(!cg->rd_io_serviced_read))
1393                 cg->rd_io_serviced_read = rrddim_add(st_io_serviced_read, cg->chart_id, cg->chart_title, 1, 1, RRDDIM_INCREMENTAL);
1394
1395             rrddim_set_by_pointer(st_io_serviced_read, cg->rd_io_serviced_read, cg->io_serviced.Read);
1396
1397             if(unlikely(!cg->rd_io_serviced_write))
1398                 cg->rd_io_serviced_write = rrddim_add(st_io_serviced_write, cg->chart_id, cg->chart_title, 1, 1, RRDDIM_INCREMENTAL);
1399
1400             rrddim_set_by_pointer(st_io_serviced_write, cg->rd_io_serviced_write, cg->io_serviced.Write);
1401         }
1402
1403         if(do_throttle_io && cg->throttle_io_service_bytes.updated) {
1404             if(unlikely(!cg->rd_throttle_io_read))
1405                 cg->rd_throttle_io_read = rrddim_add(st_throttle_io_read, cg->chart_id, cg->chart_title, 1, 1024, RRDDIM_INCREMENTAL);
1406
1407             rrddim_set_by_pointer(st_throttle_io_read, cg->rd_throttle_io_read, cg->throttle_io_service_bytes.Read);
1408
1409             if(unlikely(!cg->rd_throttle_io_write))
1410                 cg->rd_throttle_io_write = rrddim_add(st_throttle_io_write, cg->chart_id, cg->chart_title, 1, 1024, RRDDIM_INCREMENTAL);
1411
1412             rrddim_set_by_pointer(st_throttle_io_write, cg->rd_throttle_io_write, cg->throttle_io_service_bytes.Write);
1413         }
1414
1415         if(do_throttle_ops && cg->throttle_io_serviced.updated) {
1416             if(unlikely(!cg->rd_throttle_io_serviced_read))
1417                 cg->rd_throttle_io_serviced_read = rrddim_add(st_throttle_ops_read, cg->chart_id, cg->chart_title, 1, 1, RRDDIM_INCREMENTAL);
1418
1419             rrddim_set_by_pointer(st_throttle_ops_read, cg->rd_throttle_io_serviced_read, cg->throttle_io_serviced.Read);
1420
1421             if(unlikely(!cg->rd_throttle_io_serviced_write))
1422                 cg->rd_throttle_io_serviced_write = rrddim_add(st_throttle_ops_write, cg->chart_id, cg->chart_title, 1, 1, RRDDIM_INCREMENTAL);
1423
1424             rrddim_set_by_pointer(st_throttle_ops_write, cg->rd_throttle_io_serviced_write, cg->throttle_io_serviced.Write);
1425         }
1426     }
1427
1428     // complete the iteration
1429     if(do_cpu) rrdset_done(st_cpu);
1430     if(do_mem_usage) rrdset_done(st_mem_usage);
1431     if(do_io) { rrdset_done(st_io_read); rrdset_done(st_io_write); }
1432     if(do_io_ops) { rrdset_done(st_io_serviced_read); rrdset_done(st_io_serviced_write); }
1433     if(do_throttle_io) { rrdset_done(st_throttle_io_read); rrdset_done(st_throttle_io_write); }
1434     if(do_throttle_ops) { rrdset_done(st_throttle_ops_read); rrdset_done(st_throttle_ops_write); }
1435 }
1436
1437 static inline char *cgroup_chart_type(char *buffer, const char *id, size_t len) {
1438     if(buffer[0]) return buffer;
1439
1440     if(id[0] == '\0' || (id[0] == '/' && id[1] == '\0'))
1441         strncpy(buffer, "cgroup_root", len);
1442     else
1443         snprintfz(buffer, len, "cgroup_%s", id);
1444
1445     netdata_fix_chart_id(buffer);
1446     return buffer;
1447 }
1448
1449 void update_cgroup_charts(int update_every) {
1450     debug(D_CGROUP, "updating cgroups charts");
1451
1452     char type[RRD_ID_LENGTH_MAX + 1];
1453     char title[CHART_TITLE_MAX + 1];
1454
1455     int services_do_cpu = 0,
1456             services_do_mem_usage = 0,
1457             services_do_io = 0,
1458             services_do_io_ops = 0,
1459             services_do_throttle_io = 0,
1460             services_do_throttle_ops = 0,
1461             services_do_queued_ops = 0,
1462             services_do_merged_ops = 0;
1463
1464     struct cgroup *cg;
1465     for(cg = cgroup_root; cg ; cg = cg->next) {
1466         if(!cg->available || !cg->enabled) continue;
1467
1468         if(cgroup_enable_systemd_services && cg->options & CGROUP_OPTIONS_SYSTEM_SLICE_SERVICE) {
1469             if(cg->cpuacct_stat.updated && (cg->cpuacct_stat.user || cg->cpuacct_stat.system)) services_do_cpu++;
1470             if(cg->memory.usage_in_bytes_updated && (cg->memory.usage_in_bytes)) services_do_mem_usage++;
1471             if(cg->io_service_bytes.updated && (cg->io_service_bytes.Read || cg->io_service_bytes.Write)) services_do_io++;
1472             if(cg->io_serviced.updated && (cg->io_serviced.Read || cg->io_serviced.Write)) services_do_io_ops++;
1473             if(cg->throttle_io_service_bytes.updated && (cg->throttle_io_service_bytes.Read || cg->throttle_io_service_bytes.Write)) services_do_throttle_io++;
1474             if(cg->throttle_io_serviced.updated && (cg->throttle_io_serviced.Read || cg->throttle_io_serviced.Write)) services_do_throttle_ops++;
1475             if(cg->io_queued.updated && (cg->io_queued.Read || cg->io_queued.Write)) services_do_queued_ops++;
1476             if(cg->io_merged.updated && (cg->io_merged.Read || cg->io_merged.Write)) services_do_merged_ops++;
1477             continue;
1478         }
1479
1480         type[0] = '\0';
1481
1482         if(cg->cpuacct_stat.updated) {
1483             if(unlikely(!cg->st_cpu)) {
1484                 cg->st_cpu = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "cpu");
1485                 if(!cg->st_cpu) {
1486                     snprintfz(title, CHART_TITLE_MAX, "CPU Usage (%d%% = %d core%s) for cgroup %s", (processors * 100), processors, (processors > 1) ? "s" : "", cg->chart_title);
1487                     cg->st_cpu = rrdset_create(type, "cpu", NULL, "cpu", "cgroup.cpu", title, "%", 40000, update_every, RRDSET_TYPE_STACKED);
1488
1489                     rrddim_add(cg->st_cpu, "user", NULL, 100, hz, RRDDIM_INCREMENTAL);
1490                     rrddim_add(cg->st_cpu, "system", NULL, 100, hz, RRDDIM_INCREMENTAL);
1491                 }
1492             }
1493             else rrdset_next(cg->st_cpu);
1494
1495             rrddim_set(cg->st_cpu, "user", cg->cpuacct_stat.user);
1496             rrddim_set(cg->st_cpu, "system", cg->cpuacct_stat.system);
1497             rrdset_done(cg->st_cpu);
1498         }
1499
1500         if(cg->cpuacct_usage.updated) {
1501             char id[RRD_ID_LENGTH_MAX + 1];
1502             unsigned int i;
1503
1504             if(unlikely(!cg->st_cpu_per_core)) {
1505                 cg->st_cpu_per_core = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "cpu_per_core");
1506                 if(!cg->st_cpu_per_core) {
1507                     snprintfz(title, CHART_TITLE_MAX, "CPU Usage (%d%% = %d core%s) Per Core for cgroup %s", (processors * 100), processors, (processors > 1) ? "s" : "", cg->chart_title);
1508                     cg->st_cpu_per_core = rrdset_create(type, "cpu_per_core", NULL, "cpu", "cgroup.cpu_per_core", title, "%", 40100, update_every, RRDSET_TYPE_STACKED);
1509
1510                     for(i = 0; i < cg->cpuacct_usage.cpus; i++) {
1511                         snprintfz(id, CHART_TITLE_MAX, "cpu%u", i);
1512                         rrddim_add(cg->st_cpu_per_core, id, NULL, 100, 1000000000, RRDDIM_INCREMENTAL);
1513                     }
1514                 }
1515             }
1516             else rrdset_next(cg->st_cpu_per_core);
1517
1518             for(i = 0; i < cg->cpuacct_usage.cpus ;i++) {
1519                 snprintfz(id, CHART_TITLE_MAX, "cpu%u", i);
1520                 rrddim_set(cg->st_cpu_per_core, id, cg->cpuacct_usage.cpu_percpu[i]);
1521             }
1522             rrdset_done(cg->st_cpu_per_core);
1523         }
1524
1525         if(cg->memory.updated) {
1526             if(cg->memory.cache || cg->memory.rss || cg->memory.rss_huge || cg->memory.mapped_file) {
1527                 if(unlikely(!cg->st_mem)) {
1528                     cg->st_mem = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "mem");
1529                     if(!cg->st_mem) {
1530                         snprintfz(title, CHART_TITLE_MAX, "Memory Usage for cgroup %s", cg->chart_title);
1531                         cg->st_mem = rrdset_create(type, "mem", NULL, "mem", "cgroup.mem", title, "MB", 40210, update_every, RRDSET_TYPE_STACKED);
1532
1533                         rrddim_add(cg->st_mem, "cache", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1534                         rrddim_add(cg->st_mem, "rss", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1535                         if(cg->memory.has_dirty_swap)
1536                             rrddim_add(cg->st_mem, "swap", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1537                         rrddim_add(cg->st_mem, "rss_huge", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1538                         rrddim_add(cg->st_mem, "mapped_file", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1539                     }
1540                 }
1541                 else rrdset_next(cg->st_mem);
1542
1543                 rrddim_set(cg->st_mem, "cache", cg->memory.cache);
1544                 rrddim_set(cg->st_mem, "rss", cg->memory.rss);
1545                 if(cg->memory.has_dirty_swap)
1546                     rrddim_set(cg->st_mem, "swap", cg->memory.swap);
1547                 rrddim_set(cg->st_mem, "rss_huge", cg->memory.rss_huge);
1548                 rrddim_set(cg->st_mem, "mapped_file", cg->memory.mapped_file);
1549                 rrdset_done(cg->st_mem);
1550             }
1551
1552             if(unlikely(!cg->st_writeback)) {
1553                 cg->st_writeback = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "writeback");
1554                 if(!cg->st_writeback) {
1555                     snprintfz(title, CHART_TITLE_MAX, "Writeback Memory for cgroup %s", cg->chart_title);
1556                     cg->st_writeback = rrdset_create(type, "writeback", NULL, "mem", "cgroup.writeback", title, "MB", 40300, update_every, RRDSET_TYPE_AREA);
1557
1558                     if(cg->memory.has_dirty_swap)
1559                         rrddim_add(cg->st_writeback, "dirty", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1560                     rrddim_add(cg->st_writeback, "writeback", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1561                 }
1562             }
1563             else rrdset_next(cg->st_writeback);
1564
1565             if(cg->memory.has_dirty_swap)
1566                 rrddim_set(cg->st_writeback, "dirty", cg->memory.dirty);
1567             rrddim_set(cg->st_writeback, "writeback", cg->memory.writeback);
1568             rrdset_done(cg->st_writeback);
1569
1570             if(cg->memory.pgpgin || cg->memory.pgpgout) {
1571                 if(unlikely(!cg->st_mem_activity)) {
1572                     cg->st_mem_activity = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "mem_activity");
1573                     if(!cg->st_mem_activity) {
1574                         snprintfz(title, CHART_TITLE_MAX, "Memory Activity for cgroup %s", cg->chart_title);
1575                         cg->st_mem_activity = rrdset_create(type, "mem_activity", NULL, "mem", "cgroup.mem_activity", title, "MB/s", 40400, update_every, RRDSET_TYPE_LINE);
1576
1577                         rrddim_add(cg->st_mem_activity, "pgpgin", "in", sysconf(_SC_PAGESIZE), 1024 * 1024, RRDDIM_INCREMENTAL);
1578                         rrddim_add(cg->st_mem_activity, "pgpgout", "out", -sysconf(_SC_PAGESIZE), 1024 * 1024, RRDDIM_INCREMENTAL);
1579                     }
1580                 }
1581                 else rrdset_next(cg->st_mem_activity);
1582
1583                 rrddim_set(cg->st_mem_activity, "pgpgin", cg->memory.pgpgin);
1584                 rrddim_set(cg->st_mem_activity, "pgpgout", cg->memory.pgpgout);
1585                 rrdset_done(cg->st_mem_activity);
1586             }
1587
1588             if(cg->memory.pgfault || cg->memory.pgmajfault) {
1589                 if(unlikely(!cg->st_pgfaults)) {
1590                     cg->st_pgfaults = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "pgfaults");
1591                     if(!cg->st_pgfaults) {
1592                         snprintfz(title, CHART_TITLE_MAX, "Memory Page Faults for cgroup %s", cg->chart_title);
1593                         cg->st_pgfaults = rrdset_create(type, "pgfaults", NULL, "mem", "cgroup.pgfaults", title, "MB/s", 40500, update_every, RRDSET_TYPE_LINE);
1594
1595                         rrddim_add(cg->st_pgfaults, "pgfault", NULL, sysconf(_SC_PAGESIZE), 1024 * 1024, RRDDIM_INCREMENTAL);
1596                         rrddim_add(cg->st_pgfaults, "pgmajfault", "swap", -sysconf(_SC_PAGESIZE), 1024 * 1024, RRDDIM_INCREMENTAL);
1597                     }
1598                 }
1599                 else rrdset_next(cg->st_pgfaults);
1600
1601                 rrddim_set(cg->st_pgfaults, "pgfault", cg->memory.pgfault);
1602                 rrddim_set(cg->st_pgfaults, "pgmajfault", cg->memory.pgmajfault);
1603                 rrdset_done(cg->st_pgfaults);
1604             }
1605         }
1606
1607         if(cg->memory.usage_in_bytes_updated) {
1608             if(unlikely(!cg->st_mem_usage)) {
1609                 cg->st_mem_usage = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "mem_usage");
1610                 if(!cg->st_mem_usage) {
1611                     snprintfz(title, CHART_TITLE_MAX, "Total Memory for cgroup %s", cg->chart_title);
1612                     cg->st_mem_usage = rrdset_create(type, "mem_usage", NULL, "mem", "cgroup.mem_usage", title, "MB", 40200, update_every, RRDSET_TYPE_STACKED);
1613
1614                     rrddim_add(cg->st_mem_usage, "ram", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1615                     rrddim_add(cg->st_mem_usage, "swap", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
1616                 }
1617             }
1618             else rrdset_next(cg->st_mem_usage);
1619
1620             rrddim_set(cg->st_mem_usage, "ram", cg->memory.usage_in_bytes);
1621             rrddim_set(cg->st_mem_usage, "swap", (cg->memory.msw_usage_in_bytes > cg->memory.usage_in_bytes)?cg->memory.msw_usage_in_bytes - cg->memory.usage_in_bytes:0);
1622             rrdset_done(cg->st_mem_usage);
1623         }
1624
1625         if(cg->memory.failcnt_updated && cg->memory.failcnt) {
1626             if(unlikely(!cg->st_mem_failcnt)) {
1627                 cg->st_mem_failcnt = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "mem_failcnt");
1628                 if(!cg->st_mem_failcnt) {
1629                     snprintfz(title, CHART_TITLE_MAX, "Memory Limit Failures for cgroup %s", cg->chart_title);
1630                     cg->st_mem_failcnt = rrdset_create(type, "mem_failcnt", NULL, "mem", "cgroup.mem_failcnt", title, "MB", 40250, update_every, RRDSET_TYPE_LINE);
1631
1632                     rrddim_add(cg->st_mem_failcnt, "failures", NULL, 1, 1, RRDDIM_INCREMENTAL);
1633                 }
1634             }
1635             else rrdset_next(cg->st_mem_failcnt);
1636
1637             rrddim_set(cg->st_mem_failcnt, "failures", cg->memory.failcnt);
1638             rrdset_done(cg->st_mem_failcnt);
1639         }
1640
1641         if(cg->io_service_bytes.updated && (cg->io_service_bytes.Read || cg->io_service_bytes.Write)) {
1642             if(unlikely(!cg->st_io)) {
1643                 cg->st_io = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "io");
1644                 if(!cg->st_io) {
1645                     snprintfz(title, CHART_TITLE_MAX, "I/O Bandwidth (all disks) for cgroup %s", cg->chart_title);
1646                     cg->st_io = rrdset_create(type, "io", NULL, "disk", "cgroup.io", title, "KB/s", 41200, update_every, RRDSET_TYPE_AREA);
1647
1648                     rrddim_add(cg->st_io, "read", NULL, 1, 1024, RRDDIM_INCREMENTAL);
1649                     rrddim_add(cg->st_io, "write", NULL, -1, 1024, RRDDIM_INCREMENTAL);
1650                 }
1651             }
1652             else rrdset_next(cg->st_io);
1653
1654             rrddim_set(cg->st_io, "read", cg->io_service_bytes.Read);
1655             rrddim_set(cg->st_io, "write", cg->io_service_bytes.Write);
1656             rrdset_done(cg->st_io);
1657         }
1658
1659         if(cg->io_serviced.updated && (cg->io_serviced.Read || cg->io_serviced.Write)) {
1660             if(unlikely(!cg->st_serviced_ops)) {
1661                 cg->st_serviced_ops = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "serviced_ops");
1662                 if(!cg->st_serviced_ops) {
1663                     snprintfz(title, CHART_TITLE_MAX, "Serviced I/O Operations (all disks) for cgroup %s"
1664                               , cg->chart_title);
1665                     cg->st_serviced_ops = rrdset_create(type, "serviced_ops", NULL, "disk", "cgroup.serviced_ops", title, "operations/s", 41200, update_every, RRDSET_TYPE_LINE);
1666
1667                     rrddim_add(cg->st_serviced_ops, "read", NULL, 1, 1, RRDDIM_INCREMENTAL);
1668                     rrddim_add(cg->st_serviced_ops, "write", NULL, -1, 1, RRDDIM_INCREMENTAL);
1669                 }
1670             }
1671             else rrdset_next(cg->st_serviced_ops);
1672
1673             rrddim_set(cg->st_serviced_ops, "read", cg->io_serviced.Read);
1674             rrddim_set(cg->st_serviced_ops, "write", cg->io_serviced.Write);
1675             rrdset_done(cg->st_serviced_ops);
1676         }
1677
1678         if(cg->throttle_io_service_bytes.updated && (cg->throttle_io_service_bytes.Read || cg->throttle_io_service_bytes.Write)) {
1679             if(unlikely(!cg->st_throttle_io)) {
1680                 cg->st_throttle_io = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "throttle_io");
1681                 if(!cg->st_throttle_io) {
1682                     snprintfz(title, CHART_TITLE_MAX, "Throttle I/O Bandwidth (all disks) for cgroup %s", cg->chart_title);
1683                     cg->st_throttle_io = rrdset_create(type, "throttle_io", NULL, "disk", "cgroup.throttle_io", title, "KB/s", 41200, update_every, RRDSET_TYPE_AREA);
1684
1685                     rrddim_add(cg->st_throttle_io, "read", NULL, 1, 1024, RRDDIM_INCREMENTAL);
1686                     rrddim_add(cg->st_throttle_io, "write", NULL, -1, 1024, RRDDIM_INCREMENTAL);
1687                 }
1688             }
1689             else rrdset_next(cg->st_throttle_io);
1690
1691             rrddim_set(cg->st_throttle_io, "read", cg->throttle_io_service_bytes.Read);
1692             rrddim_set(cg->st_throttle_io, "write", cg->throttle_io_service_bytes.Write);
1693             rrdset_done(cg->st_throttle_io);
1694         }
1695
1696         if(cg->throttle_io_serviced.updated && (cg->throttle_io_serviced.Read || cg->throttle_io_serviced.Write)) {
1697             if(unlikely(!cg->st_throttle_serviced_ops)) {
1698                 cg->st_throttle_serviced_ops = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "throttle_serviced_ops");
1699                 if(!cg->st_throttle_serviced_ops) {
1700                     snprintfz(title, CHART_TITLE_MAX, "Throttle Serviced I/O Operations (all disks) for cgroup %s", cg->chart_title);
1701                     cg->st_throttle_serviced_ops = rrdset_create(type, "throttle_serviced_ops", NULL, "disk", "cgroup.throttle_serviced_ops", title, "operations/s", 41200, update_every, RRDSET_TYPE_LINE);
1702
1703                     rrddim_add(cg->st_throttle_serviced_ops, "read", NULL, 1, 1, RRDDIM_INCREMENTAL);
1704                     rrddim_add(cg->st_throttle_serviced_ops, "write", NULL, -1, 1, RRDDIM_INCREMENTAL);
1705                 }
1706             }
1707             else rrdset_next(cg->st_throttle_serviced_ops);
1708
1709             rrddim_set(cg->st_throttle_serviced_ops, "read", cg->throttle_io_serviced.Read);
1710             rrddim_set(cg->st_throttle_serviced_ops, "write", cg->throttle_io_serviced.Write);
1711             rrdset_done(cg->st_throttle_serviced_ops);
1712         }
1713
1714         if(cg->io_queued.updated) {
1715             if(unlikely(!cg->st_queued_ops)) {
1716                 cg->st_queued_ops = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "queued_ops");
1717                 if(!cg->st_queued_ops) {
1718                     snprintfz(title, CHART_TITLE_MAX, "Queued I/O Operations (all disks) for cgroup %s", cg->chart_title);
1719                     cg->st_queued_ops = rrdset_create(type, "queued_ops", NULL, "disk", "cgroup.queued_ops", title, "operations", 42000, update_every, RRDSET_TYPE_LINE);
1720
1721                     rrddim_add(cg->st_queued_ops, "read", NULL, 1, 1, RRDDIM_ABSOLUTE);
1722                     rrddim_add(cg->st_queued_ops, "write", NULL, -1, 1, RRDDIM_ABSOLUTE);
1723                 }
1724             }
1725             else rrdset_next(cg->st_queued_ops);
1726
1727             rrddim_set(cg->st_queued_ops, "read", cg->io_queued.Read);
1728             rrddim_set(cg->st_queued_ops, "write", cg->io_queued.Write);
1729             rrdset_done(cg->st_queued_ops);
1730         }
1731
1732         if(cg->io_merged.updated && (cg->io_merged.Read || cg->io_merged.Write)) {
1733             if(unlikely(!cg->st_merged_ops)) {
1734                 cg->st_merged_ops = rrdset_find_bytype(cgroup_chart_type(type, cg->chart_id, RRD_ID_LENGTH_MAX), "merged_ops");
1735                 if(!cg->st_merged_ops) {
1736                     snprintfz(title, CHART_TITLE_MAX, "Merged I/O Operations (all disks) for cgroup %s", cg->chart_title);
1737                     cg->st_merged_ops = rrdset_create(type, "merged_ops", NULL, "disk", "cgroup.merged_ops", title, "operations/s", 42100, update_every, RRDSET_TYPE_LINE);
1738
1739                     rrddim_add(cg->st_merged_ops, "read", NULL, 1, 1024, RRDDIM_INCREMENTAL);
1740                     rrddim_add(cg->st_merged_ops, "write", NULL, -1, 1024, RRDDIM_INCREMENTAL);
1741                 }
1742             }
1743             else rrdset_next(cg->st_merged_ops);
1744
1745             rrddim_set(cg->st_merged_ops, "read", cg->io_merged.Read);
1746             rrddim_set(cg->st_merged_ops, "write", cg->io_merged.Write);
1747             rrdset_done(cg->st_merged_ops);
1748         }
1749     }
1750
1751     debug(D_CGROUP, "done updating cgroups charts");
1752
1753     if(cgroup_enable_systemd_services)
1754         update_services_charts(update_every,
1755                 services_do_cpu,
1756                 services_do_mem_usage,
1757                 services_do_io,
1758                 services_do_io_ops,
1759                 services_do_throttle_io,
1760                 services_do_throttle_ops/*,
1761                 services_do_queued_ops,
1762                 services_do_merged_ops*/
1763         );
1764 }
1765
1766 // ----------------------------------------------------------------------------
1767 // cgroups main
1768
1769 void *cgroups_main(void *ptr) {
1770     struct netdata_static_thread *static_thread = (struct netdata_static_thread *)ptr;
1771
1772     info("CGROUP Plugin thread created with task id %d", gettid());
1773
1774     if(pthread_setcanceltype(PTHREAD_CANCEL_DEFERRED, NULL) != 0)
1775         error("Cannot set pthread cancel type to DEFERRED.");
1776
1777     if(pthread_setcancelstate(PTHREAD_CANCEL_ENABLE, NULL) != 0)
1778         error("Cannot set pthread cancel state to ENABLE.");
1779
1780     struct rusage thread;
1781
1782     // when ZERO, attempt to do it
1783     int vdo_cpu_netdata = config_get_boolean("plugin:cgroups", "cgroups plugin resource charts", 1);
1784
1785     read_cgroup_plugin_configuration();
1786
1787     RRDSET *stcpu_thread = NULL;
1788
1789     usec_t step = cgroup_update_every * USEC_PER_SEC;
1790     usec_t find_every = cgroup_check_for_new_every * USEC_PER_SEC, find_next = 0;
1791     for(;;) {
1792         usec_t now = now_monotonic_usec();
1793         usec_t next = now - (now % step) + step;
1794
1795         while(now < next) {
1796             sleep_usec(next - now);
1797             now = now_monotonic_usec();
1798         }
1799
1800         if(unlikely(netdata_exit)) break;
1801
1802         // BEGIN -- the job to be done
1803
1804         if(unlikely(now >= find_next)) {
1805             find_all_cgroups();
1806             find_next = now + find_every;
1807         }
1808
1809         read_all_cgroups(cgroup_root);
1810         update_cgroup_charts(cgroup_update_every);
1811
1812         // END -- the job is done
1813
1814         // --------------------------------------------------------------------
1815
1816         if(vdo_cpu_netdata) {
1817             getrusage(RUSAGE_THREAD, &thread);
1818
1819             if(!stcpu_thread) stcpu_thread = rrdset_find("netdata.plugin_cgroups_cpu");
1820             if(!stcpu_thread) {
1821                 stcpu_thread = rrdset_create("netdata", "plugin_cgroups_cpu", NULL, "cgroups", NULL, "NetData CGroups Plugin CPU usage", "milliseconds/s", 132000, cgroup_update_every, RRDSET_TYPE_STACKED);
1822
1823                 rrddim_add(stcpu_thread, "user",  NULL,  1, 1000, RRDDIM_INCREMENTAL);
1824                 rrddim_add(stcpu_thread, "system", NULL, 1, 1000, RRDDIM_INCREMENTAL);
1825             }
1826             else rrdset_next(stcpu_thread);
1827
1828             rrddim_set(stcpu_thread, "user"  , thread.ru_utime.tv_sec * 1000000ULL + thread.ru_utime.tv_usec);
1829             rrddim_set(stcpu_thread, "system", thread.ru_stime.tv_sec * 1000000ULL + thread.ru_stime.tv_usec);
1830             rrdset_done(stcpu_thread);
1831         }
1832     }
1833
1834     info("CGROUP thread exiting");
1835
1836     static_thread->enabled = 0;
1837     pthread_exit(NULL);
1838     return NULL;
1839 }