unsigned long long total_active_file;
unsigned long long total_unevictable;
*/
+
+ int usage_in_bytes_updated;
+ char *filename_usage_in_bytes;
+ unsigned long long usage_in_bytes;
+
+ int msw_usage_in_bytes_updated;
+ char *filename_msw_usage_in_bytes;
+ unsigned long long msw_usage_in_bytes;
+
+ int failcnt_updated;
+ char *filename_failcnt;
+ unsigned long long failcnt;
};
// https://www.kernel.org/doc/Documentation/cgroup-v1/cpuacct.txt
unsigned long long *cpu_percpu;
};
+#define CGROUP_OPTIONS_DISABLED_DUPLICATE 0x00000001
+
struct cgroup {
- int available; // found in the filesystem
- int enabled; // enabled in the config
+ uint32_t options;
+
+ char available; // found in the filesystem
+ char enabled; // enabled in the config
char *id;
uint32_t hash;
}
unsigned long i = procfile_linewords(ff, 0);
- if(i <= 0) return;
+ if(i == 0) return;
// we may have 1 more CPU reported
while(i > 0) {
mem->updated = 1;
}
+
+ mem->usage_in_bytes_updated = 0;
+ if(mem->filename_usage_in_bytes) {
+ if(likely(!read_single_number_file(mem->filename_usage_in_bytes, &mem->usage_in_bytes)))
+ mem->usage_in_bytes_updated = 1;
+ }
+
+ mem->msw_usage_in_bytes_updated = 0;
+ if(mem->filename_msw_usage_in_bytes) {
+ if(likely(!read_single_number_file(mem->filename_msw_usage_in_bytes, &mem->msw_usage_in_bytes)))
+ mem->msw_usage_in_bytes_updated = 1;
+ }
+
+ mem->failcnt_updated = 0;
+ if(mem->filename_failcnt) {
+ if(likely(!read_single_number_file(mem->filename_failcnt, &mem->failcnt)))
+ mem->failcnt_updated = 1;
+ }
}
void cgroup_read(struct cgroup *cg) {
// starts with them
(len > 6 && !strncmp(chart_id, "user/", 6)) ||
(len > 11 && !strncmp(chart_id, "user.slice/", 11)) ||
- (len > 11 && !strncmp(chart_id, "init.scope/", 11)) ||
// ends with them
(len > 5 && !strncmp(&chart_id[len - 5], ".user", 5)) ||
(len > 5 && !strncmp(&chart_id[len - 5], ".swap", 5)) ||
cgroup_root_count++;
// fix the name by calling the external script
- if(def) cgroup_get_chart_id(cg);
+ cgroup_get_chart_id(cg);
debug(D_CGROUP, "adding cgroup '%s' with chart id '%s'", id, chart_id);
snprintfz(option, FILENAME_MAX, "enable cgroup %s", cg->chart_title);
cg->enabled = config_get_boolean("plugin:cgroups", option, def);
- struct cgroup *t;
- for(t = cgroup_root; t ; t = t->next) {
- if(t != cg && t->hash_chart == cg->hash_chart && !strcmp(t->chart_id, cg->chart_id) && t->enabled) {
- error("Control group with chart id '%s' already exists and is enabled. Disabling cgroup '%s'.", cg->chart_id, cg->id);
- cg->enabled = 0;
+ if(cg->enabled) {
+ struct cgroup *t;
+ for (t = cgroup_root; t; t = t->next) {
+ if (t != cg && t->enabled && t->hash_chart == cg->hash_chart && !strcmp(t->chart_id, cg->chart_id)) {
+ if (!strncmp(t->chart_id, "/system.slice/", 14) && !strncmp(cg->chart_id, "/init.scope/system.slice/", 25)) {
+ 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'.",
+ cg->chart_id, t->id, cg->id, t->id);
+ 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'.",
+ cg->chart_id, t->id, cg->id, t->id);
+ t->enabled = 0;
+ t->options |= CGROUP_OPTIONS_DISABLED_DUPLICATE;
+ }
+ else {
+ error("Control group with chart id '%s' already exists with id '%s' and is enabled and available. Disabling cgroup with id '%s'.",
+ cg->chart_id, t->id, cg->id);
+ debug(D_CGROUP, "Control group with chart id '%s' already exists with id '%s' and is enabled and available. Disabling cgroup with id '%s'.",
+ cg->chart_id, t->id, cg->id);
+ cg->enabled = 0;
+ cg->options |= CGROUP_OPTIONS_DISABLED_DUPLICATE;
+ }
+
+ break;
+ }
}
}
if(cg) cg->available = 1;
}
-void find_dir_in_subdirs(const char *base, const char *this, void (*callback)(const char *)) {
+int find_dir_in_subdirs(const char *base, const char *this, void (*callback)(const char *)) {
debug(D_CGROUP, "searching for directories in '%s'", base);
+ int ret = -1;
int enabled = -1;
if(!this) this = base;
size_t dirlen = strlen(this), baselen = strlen(base);
DIR *dir = opendir(this);
if(!dir) {
error("Cannot read cgroups directory '%s'", base);
- return;
+ return ret;
}
+ ret = 1;
callback(relative_path);
if(*r == '\0') r = "/";
else if (*r == '/') r++;
+ // do not decent in directories we are not interested
+ // https://github.com/firehol/netdata/issues/345
+ int def = 1;
+ size_t len = strlen(r);
+ if(len > 5 && !strncmp(&r[len - 5], "-qemu", 5))
+ def = 0;
+
// we check for this option here
// so that the config will not have settings
// for leaf directories
char option[FILENAME_MAX + 1];
snprintfz(option, FILENAME_MAX, "search for cgroups under %s", r);
option[FILENAME_MAX] = '\0';
- enabled = config_get_boolean("plugin:cgroups", option, 1);
+ enabled = config_get_boolean("plugin:cgroups", option, def);
}
if(enabled) {
strcpy(s, this);
strcat(s, "/");
strcat(s, de->d_name);
- find_dir_in_subdirs(base, s, callback);
+ int ret2 = find_dir_in_subdirs(base, s, callback);
+ if(ret2 > 0) ret += ret2;
freez(s);
}
}
}
closedir(dir);
+ return ret;
}
void mark_all_cgroups_as_not_available() {
struct cgroup *cg;
// mark all as not available
- for(cg = cgroup_root; cg ; cg = cg->next)
+ for(cg = cgroup_root; cg ; cg = cg->next) {
cg->available = 0;
+ }
}
void cleanup_all_cgroups() {
for(; cg ;) {
if(!cg->available) {
+ // enable the first duplicate cgroup
+ {
+ struct cgroup *t;
+ for(t = cgroup_root; t ; t = t->next) {
+ 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)) {
+ 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);
+ t->enabled = 1;
+ t->options &= ~CGROUP_OPTIONS_DISABLED_DUPLICATE;
+ break;
+ }
+ }
+ }
if(!last)
cgroup_root = cg->next;
mark_all_cgroups_as_not_available();
- if(cgroup_enable_cpuacct_stat || cgroup_enable_cpuacct_usage)
- find_dir_in_subdirs(cgroup_cpuacct_base, NULL, found_subdir_in_dir);
+ if(cgroup_enable_cpuacct_stat || cgroup_enable_cpuacct_usage) {
+ if (find_dir_in_subdirs(cgroup_cpuacct_base, NULL, found_subdir_in_dir) == -1) {
+ cgroup_enable_cpuacct_stat = cgroup_enable_cpuacct_usage = 0;
+ error("disabled cgroup cpu statistics.");
+ }
+ }
- if(cgroup_enable_blkio)
- find_dir_in_subdirs(cgroup_blkio_base, NULL, found_subdir_in_dir);
+ if(cgroup_enable_blkio) {
+ if (find_dir_in_subdirs(cgroup_blkio_base, NULL, found_subdir_in_dir) == -1) {
+ cgroup_enable_blkio = 0;
+ error("disabled cgroup blkio statistics.");
+ }
+ }
- if(cgroup_enable_memory)
- find_dir_in_subdirs(cgroup_memory_base, NULL, found_subdir_in_dir);
+ if(cgroup_enable_memory) {
+ if(find_dir_in_subdirs(cgroup_memory_base, NULL, found_subdir_in_dir) == -1) {
+ cgroup_enable_memory = 0;
+ error("disabled cgroup memory statistics.");
+ }
+ }
- if(cgroup_enable_devices)
- find_dir_in_subdirs(cgroup_devices_base, NULL, found_subdir_in_dir);
+ if(cgroup_enable_devices) {
+ if(find_dir_in_subdirs(cgroup_devices_base, NULL, found_subdir_in_dir) == -1) {
+ cgroup_enable_devices = 0;
+ error("disabled cgroup devices statistics.");
+ }
+ }
// remove any non-existing cgroups
cleanup_all_cgroups();
debug(D_CGROUP, "memory.stat filename for cgroup '%s': '%s'", cg->id, cg->memory.filename);
}
else debug(D_CGROUP, "memory.stat file for cgroup '%s': '%s' does not exist.", cg->id, filename);
+
+ snprintfz(filename, FILENAME_MAX, "%s%s/memory.usage_in_bytes", cgroup_memory_base, cg->id);
+ if(stat(filename, &buf) != -1) {
+ cg->memory.filename_usage_in_bytes = strdupz(filename);
+ debug(D_CGROUP, "memory.usage_in_bytes filename for cgroup '%s': '%s'", cg->id, cg->memory.filename_usage_in_bytes);
+ }
+ else debug(D_CGROUP, "memory.usage_in_bytes file for cgroup '%s': '%s' does not exist.", cg->id, filename);
+
+ snprintfz(filename, FILENAME_MAX, "%s%s/memory.msw_usage_in_bytes", cgroup_memory_base, cg->id);
+ if(stat(filename, &buf) != -1) {
+ cg->memory.filename_msw_usage_in_bytes = strdupz(filename);
+ debug(D_CGROUP, "memory.msw_usage_in_bytes filename for cgroup '%s': '%s'", cg->id, cg->memory.filename_msw_usage_in_bytes);
+ }
+ else debug(D_CGROUP, "memory.msw_usage_in_bytes file for cgroup '%s': '%s' does not exist.", cg->id, filename);
+
+ snprintfz(filename, FILENAME_MAX, "%s%s/memory.failcnt", cgroup_memory_base, cg->id);
+ if(stat(filename, &buf) != -1) {
+ cg->memory.filename_failcnt = strdupz(filename);
+ debug(D_CGROUP, "memory.failcnt filename for cgroup '%s': '%s'", cg->id, cg->memory.filename_failcnt);
+ }
+ else debug(D_CGROUP, "memory.failcnt file for cgroup '%s': '%s' does not exist.", cg->id, filename);
}
if(cgroup_enable_blkio) {
if(!cg->io_service_bytes.filename) {
if(cg->cpuacct_stat.updated) {
st = rrdset_find_bytype(type, "cpu");
if(!st) {
- snprintfz(title, CHART_TITLE_MAX, "CPU Usage for cgroup %s", cg->chart_title);
+ snprintfz(title, CHART_TITLE_MAX, "CPU Usage (%d%% = %d core%s) for cgroup %s", (processors * 100), processors, (processors>1)?"s":"", cg->chart_title);
st = rrdset_create(type, "cpu", NULL, "cpu", "cgroup.cpu", title, "%", 40000, update_every, RRDSET_TYPE_STACKED);
rrddim_add(st, "user", NULL, 100, hz, RRDDIM_INCREMENTAL);
st = rrdset_find_bytype(type, "cpu_per_core");
if(!st) {
- snprintfz(title, CHART_TITLE_MAX, "CPU Usage Per Core for cgroup %s", cg->chart_title);
+ 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);
st = rrdset_create(type, "cpu_per_core", NULL, "cpu", "cgroup.cpu_per_core", title, "%", 40100, update_every, RRDSET_TYPE_STACKED);
for(i = 0; i < cg->cpuacct_usage.cpus ;i++) {
st = rrdset_find_bytype(type, "mem");
if(!st) {
snprintfz(title, CHART_TITLE_MAX, "Memory Usage for cgroup %s", cg->chart_title);
- st = rrdset_create(type, "mem", NULL, "mem", "cgroup.mem", title, "MB", 40200, update_every,
+ st = rrdset_create(type, "mem", NULL, "mem", "cgroup.mem", title, "MB", 40210, update_every,
RRDSET_TYPE_STACKED);
rrddim_add(st, "cache", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
}
}
+ if(cg->memory.usage_in_bytes_updated) {
+ st = rrdset_find_bytype(type, "mem_usage");
+ if(!st) {
+ snprintfz(title, CHART_TITLE_MAX, "Total Memory for cgroup %s", cg->chart_title);
+ st = rrdset_create(type, "mem_usage", NULL, "mem", "cgroup.mem_usage", title, "MB", 40200,
+ update_every, RRDSET_TYPE_STACKED);
+
+ rrddim_add(st, "ram", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
+ rrddim_add(st, "swap", NULL, 1, 1024 * 1024, RRDDIM_ABSOLUTE);
+ }
+ else rrdset_next(st);
+
+ rrddim_set(st, "ram", cg->memory.usage_in_bytes);
+ rrddim_set(st, "swap", (cg->memory.msw_usage_in_bytes > cg->memory.usage_in_bytes)?cg->memory.msw_usage_in_bytes - cg->memory.usage_in_bytes:0);
+ rrdset_done(st);
+ }
+
+ if(cg->memory.failcnt_updated && cg->memory.failcnt > 0) {
+ st = rrdset_find_bytype(type, "mem_failcnt");
+ if(!st) {
+ snprintfz(title, CHART_TITLE_MAX, "Memory Limit Failures for cgroup %s", cg->chart_title);
+ st = rrdset_create(type, "mem_failcnt", NULL, "mem", "cgroup.mem_failcnt", title, "MB", 40250,
+ update_every, RRDSET_TYPE_LINE);
+
+ rrddim_add(st, "failures", NULL, 1, 1, RRDDIM_INCREMENTAL);
+ }
+ else rrdset_next(st);
+
+ rrddim_set(st, "failures", cg->memory.failcnt);
+ rrdset_done(st);
+ }
+
if(cg->io_service_bytes.updated && cg->io_service_bytes.Read + cg->io_service_bytes.Write > 0) {
st = rrdset_find_bytype(type, "io");
if(!st) {
snprintfz(title, CHART_TITLE_MAX, "I/O Bandwidth (all disks) for cgroup %s", cg->chart_title);
st = rrdset_create(type, "io", NULL, "disk", "cgroup.io", title, "KB/s", 41200,
- update_every, RRDSET_TYPE_LINE);
+ update_every, RRDSET_TYPE_AREA);
rrddim_add(st, "read", NULL, 1, 1024, RRDDIM_INCREMENTAL);
rrddim_add(st, "write", NULL, -1, 1024, RRDDIM_INCREMENTAL);
}
if(cg->throttle_io_service_bytes.updated && cg->throttle_io_service_bytes.Read + cg->throttle_io_service_bytes.Write > 0) {
- st = rrdset_find_bytype(type, "io");
+ st = rrdset_find_bytype(type, "throttle_io");
if(!st) {
snprintfz(title, CHART_TITLE_MAX, "Throttle I/O Bandwidth (all disks) for cgroup %s", cg->chart_title);
- st = rrdset_create(type, "io", NULL, "disk", "cgroup.io", title, "KB/s", 41200,
- update_every, RRDSET_TYPE_LINE);
+ st = rrdset_create(type, "throttle_io", NULL, "disk", "cgroup.throttle_io", title, "KB/s", 41200,
+ update_every, RRDSET_TYPE_AREA);
rrddim_add(st, "read", NULL, 1, 1024, RRDDIM_INCREMENTAL);
rrddim_add(st, "write", NULL, -1, 1024, RRDDIM_INCREMENTAL);
// ----------------------------------------------------------------------------
// cgroups main
-int do_sys_fs_cgroup(int update_every, unsigned long long dt) {
+int do_sys_fs_cgroup(int update_every, usec_t dt) {
(void)dt;
static int cgroup_global_config_read = 0;
static time_t last_run = 0;
- time_t now = time(NULL);
+ time_t now = now_realtime_sec();
if(unlikely(!cgroup_global_config_read)) {
read_cgroup_plugin_configuration();
void *cgroups_main(void *ptr)
{
- if(ptr) { ; }
+ (void)ptr;
info("CGROUP Plugin thread created with task id %d", gettid());
int vdo_cpu_netdata = !config_get_boolean("plugin:cgroups", "cgroups plugin resources", 1);
// keep track of the time each module was called
- unsigned long long sutime_sys_fs_cgroup = 0ULL;
+ usec_t sutime_sys_fs_cgroup = 0ULL;
// the next time we will run - aligned properly
- unsigned long long sunext = (time(NULL) - (time(NULL) % rrd_update_every) + rrd_update_every) * 1000000ULL;
- unsigned long long sunow;
+ usec_t sunext = (now_realtime_sec() - (now_realtime_sec() % rrd_update_every) + rrd_update_every) * USEC_PER_SEC;
RRDSET *stcpu_thread = NULL;
- for(;1;) {
+ for(;;) {
+ usec_t sunow;
if(unlikely(netdata_exit)) break;
// delay until it is our time to run
- while((sunow = time_usec()) < sunext)
+ while((sunow = now_realtime_usec()) < sunext)
sleep_usec(sunext - sunow);
// find the next time we need to run
- while(time_usec() > sunext)
- sunext += rrd_update_every * 1000000ULL;
+ while(now_realtime_usec() > sunext)
+ sunext += rrd_update_every * USEC_PER_SEC;
if(unlikely(netdata_exit)) break;
if(!vdo_sys_fs_cgroup) {
debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_sys_fs_cgroup().");
- sunow = time_usec();
+ sunow = now_realtime_usec();
vdo_sys_fs_cgroup = do_sys_fs_cgroup(rrd_update_every, (sutime_sys_fs_cgroup > 0)?sunow - sutime_sys_fs_cgroup:0ULL);
sutime_sys_fs_cgroup = sunow;
}
}
}
+ info("CGROUP thread exiting");
+
pthread_exit(NULL);
return NULL;
}
projects / netdata.git / blobdiff