cgroup_enable_blkio = config_get_boolean_ondemand("plugin:cgroups", "enable blkio", cgroup_enable_blkio);
char filename[FILENAME_MAX + 1], *s;
- struct mountinfo *mi, *root = mountinfo_read();
+ struct mountinfo *mi, *root = mountinfo_read(0);
mi = mountinfo_find_by_filesystem_super_option(root, "cgroup", "cpuacct");
if(!mi) mi = mountinfo_find_by_filesystem_mount_source(root, "cgroup", "cpuacct");
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) {
// disable by default the root cgroup
def = 0;
- debug(D_CGROUP, "cgroup '%s' is the root container (by default %s)", id, (def)?"enabled":"disabled");
}
else {
if(*chart_id == '/') chart_id++;
!strcmp(chart_id, "system") ||
!strcmp(chart_id, "machine") ||
// starts with them
- (len > 6 && !strncmp(chart_id, "user/", 6)) ||
+ (len > 5 && !strncmp(chart_id, "user/", 5)) ||
(len > 11 && !strncmp(chart_id, "user.slice/", 11)) ||
// ends with them
(len > 5 && !strncmp(&chart_id[len - 5], ".user", 5)) ||
(len > 6 && !strncmp(&chart_id[len - 6], ".mount", 6)) ||
(len > 8 && !strncmp(&chart_id[len - 8], ".session", 8)) ||
(len > 8 && !strncmp(&chart_id[len - 8], ".service", 8)) ||
- (len > 10 && !strncmp(&chart_id[len - 10], ".partition", 10))
+ (len > 10 && !strncmp(&chart_id[len - 10], ".partition", 10)) ||
+ // starts and ends with them
+ (len > 7 && !strncmp(chart_id, "lxc/", 4) && !strncmp(&chart_id[len - 3], "/ns", 3)) // #1397
) {
def = 0;
- debug(D_CGROUP, "cgroup '%s' is %s (by default)", id, (def)?"enabled":"disabled");
}
}
+ debug(D_CGROUP, "cgroup '%s' (chart_id '%s') is (by default) %s", id, chart_id, (def)?"enabled":"disabled");
struct cgroup *cg = callocz(1, sizeof(struct cgroup));
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;
- } else {
- error("Control group with chart id '%s' already exists with id '%s' and is enabled. Disabling cgroup with id '%s'.",
+ 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;
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 = 0;
- int enabled = 0;
+ int ret = -1;
+ int enabled = -1;
if(!this) this = base;
size_t dirlen = strlen(this), baselen = strlen(base);
const char *relative_path = &this[baselen];
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) {
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;
// check for newly added cgroups
// and update the filenames they read
char filename[FILENAME_MAX + 1];
- if(cgroup_enable_cpuacct_stat && !cg->cpuacct_stat.filename) {
+ if(unlikely(cgroup_enable_cpuacct_stat && !cg->cpuacct_stat.filename)) {
snprintfz(filename, FILENAME_MAX, "%s%s/cpuacct.stat", cgroup_cpuacct_base, cg->id);
if(stat(filename, &buf) != -1) {
cg->cpuacct_stat.filename = strdupz(filename);
}
else debug(D_CGROUP, "cpuacct.stat file for cgroup '%s': '%s' does not exist.", cg->id, filename);
}
- if(cgroup_enable_cpuacct_usage && !cg->cpuacct_usage.filename) {
+
+ if(unlikely(cgroup_enable_cpuacct_usage && !cg->cpuacct_usage.filename)) {
snprintfz(filename, FILENAME_MAX, "%s%s/cpuacct.usage_percpu", cgroup_cpuacct_base, cg->id);
if(stat(filename, &buf) != -1) {
cg->cpuacct_usage.filename = strdupz(filename);
}
else debug(D_CGROUP, "cpuacct.usage_percpu file for cgroup '%s': '%s' does not exist.", cg->id, filename);
}
- if(cgroup_enable_memory && !cg->memory.filename) {
- snprintfz(filename, FILENAME_MAX, "%s%s/memory.stat", cgroup_memory_base, cg->id);
- if(stat(filename, &buf) != -1) {
- cg->memory.filename = strdupz(filename);
- debug(D_CGROUP, "memory.stat filename for cgroup '%s': '%s'", cg->id, cg->memory.filename);
+
+ if(unlikely(cgroup_enable_memory)) {
+ if(unlikely(!cg->memory.filename)) {
+ snprintfz(filename, FILENAME_MAX, "%s%s/memory.stat", cgroup_memory_base, cg->id);
+ if(stat(filename, &buf) != -1) {
+ cg->memory.filename = strdupz(filename);
+ 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);
+ }
+ if(unlikely(!cg->memory.filename_usage_in_bytes)) {
+ 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);
+ }
+ if(unlikely(!cg->memory.filename_msw_usage_in_bytes)) {
+ 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);
+ }
+ if(unlikely(!cg->memory.filename_failcnt)) {
+ 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);
}
- else debug(D_CGROUP, "memory.stat file for cgroup '%s': '%s' does not exist.", cg->id, filename);
}
- if(cgroup_enable_blkio) {
- if(!cg->io_service_bytes.filename) {
+
+ if(unlikely(cgroup_enable_blkio)) {
+ if(unlikely(!cg->io_service_bytes.filename)) {
snprintfz(filename, FILENAME_MAX, "%s%s/blkio.io_service_bytes", cgroup_blkio_base, cg->id);
if(stat(filename, &buf) != -1) {
cg->io_service_bytes.filename = strdupz(filename);
}
else debug(D_CGROUP, "io_service_bytes file for cgroup '%s': '%s' does not exist.", cg->id, filename);
}
- if(!cg->io_serviced.filename) {
+ if(unlikely(!cg->io_serviced.filename)) {
snprintfz(filename, FILENAME_MAX, "%s%s/blkio.io_serviced", cgroup_blkio_base, cg->id);
if(stat(filename, &buf) != -1) {
cg->io_serviced.filename = strdupz(filename);
}
else debug(D_CGROUP, "io_serviced file for cgroup '%s': '%s' does not exist.", cg->id, filename);
}
- if(!cg->throttle_io_service_bytes.filename) {
+ if(unlikely(!cg->throttle_io_service_bytes.filename)) {
snprintfz(filename, FILENAME_MAX, "%s%s/blkio.throttle.io_service_bytes", cgroup_blkio_base, cg->id);
if(stat(filename, &buf) != -1) {
cg->throttle_io_service_bytes.filename = strdupz(filename);
}
else debug(D_CGROUP, "throttle_io_service_bytes file for cgroup '%s': '%s' does not exist.", cg->id, filename);
}
- if(!cg->throttle_io_serviced.filename) {
+ if(unlikely(!cg->throttle_io_serviced.filename)) {
snprintfz(filename, FILENAME_MAX, "%s%s/blkio.throttle.io_serviced", cgroup_blkio_base, cg->id);
if(stat(filename, &buf) != -1) {
cg->throttle_io_serviced.filename = strdupz(filename);
}
else debug(D_CGROUP, "throttle_io_serviced file for cgroup '%s': '%s' does not exist.", cg->id, filename);
}
- if(!cg->io_merged.filename) {
+ if(unlikely(!cg->io_merged.filename)) {
snprintfz(filename, FILENAME_MAX, "%s%s/blkio.io_merged", cgroup_blkio_base, cg->id);
if(stat(filename, &buf) != -1) {
cg->io_merged.filename = strdupz(filename);
}
else debug(D_CGROUP, "io_merged file for cgroup '%s': '%s' does not exist.", cg->id, filename);
}
- if(!cg->io_queued.filename) {
+ if(unlikely(!cg->io_queued.filename)) {
snprintfz(filename, FILENAME_MAX, "%s%s/blkio.io_queued", cgroup_blkio_base, cg->id);
if(stat(filename, &buf) != -1) {
cg->io_queued.filename = strdupz(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();
return 0;
}
-void *cgroups_main(void *ptr)
-{
- if(ptr) { ; }
+void *cgroups_main(void *ptr) {
+ struct netdata_static_thread *static_thread = (struct netdata_static_thread *)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;
}
if(!stcpu_thread) stcpu_thread = rrdset_find("netdata.plugin_cgroups_cpu");
if(!stcpu_thread) {
- stcpu_thread = rrdset_create("netdata", "plugin_cgroups_cpu", NULL, "proc.internal", NULL, "NetData CGroups Plugin CPU usage", "milliseconds/s", 132000, rrd_update_every, RRDSET_TYPE_STACKED);
+ stcpu_thread = rrdset_create("netdata", "plugin_cgroups_cpu", NULL, "cgroups", NULL, "NetData CGroups Plugin CPU usage", "milliseconds/s", 132000, rrd_update_every, RRDSET_TYPE_STACKED);
rrddim_add(stcpu_thread, "user", NULL, 1, 1000, RRDDIM_INCREMENTAL);
rrddim_add(stcpu_thread, "system", NULL, 1, 1000, RRDDIM_INCREMENTAL);
}
}
+ info("CGROUP thread exiting");
+
+ static_thread->enabled = 0;
+ static_thread->thread = NULL;
pthread_exit(NULL);
return NULL;
}
projects / netdata.git / blobdiff