@cryptotaxi247 / netdata-1 / commits / b1dfe19b4

cgroups are now on their own thread; cgroup names are controlled by plugins.d/cgroup-name.sh which uses /etc/netdata/cgroups-names.conf to pick up unique ID (chart names) for the containers

Costa Tsaousis (ktsaou) committed Apr 25, 2016 at 01:33 UTC b1dfe19b4f139a7e213d078dab8dd430040851b2

9 files changed +182 -47

plugins.d/Makefile.am

@@ -8,6 +8,7 @@ dist_plugins_DATA = \

8 $(NULL)

10 dist_plugins_SCRIPTS = \

11 + cgroup-name.sh \

12 charts.d.dryrun-helper.sh \

13 charts.d.plugin \

14 node.d.plugin \

plugins.d/cgroup-name.sh new

+38

@@ -0,0 +1,38 @@

1 +#!/bin/bash

2 +

3 +export PATH="${PATH}:/sbin:/usr/sbin:/usr/local/sbin"

4 +

5 +NETDATA_CONFIG_DIR="${NETDATA_CONFIG_DIR-/etc/netdata}"

6 +CONFIG="${NETDATA_CONFIG_DIR}/cgroups-names.conf"

7 +CGROUP="${1}"

8 +NAME=

9 +

10 +if [ -z "${CGROUP}" ]

11 + then

12 + echo >&2 "${0}: called without a cgroup name. Nothing to do."

13 + exit 1

14 +fi

15 +

16 +if [ -f "${CONFIG}" ]

17 + then

18 + NAME="$(cat "${CONFIG}" | grep "^${CGROUP}" | cut -d ' ' -f 2)"

19 + if [ -z "${NAME}" ]

20 + then

21 + echo >&2 "${0}: cannot find cgroup '${CGROUP}' in '${CONFIG}'."

22 + fi

23 +else

24 + echo >&2 "${0}: configuration file '${CONFIG}' is not available."

25 +fi

26 +

27 +if [ -z "${NAME}" ]

28 + then

29 + if [ ${#CGROUP} -gt 12 ]

30 + then

31 + NAME="${CGROUP:0:12}"

32 + else

33 + NAME="${CGROUP}"

34 + fi

35 +fi

36 +

37 +echo >&2 "${0}: cgroup '${CGROUP}' is named as '${NAME}'"

38 +echo "${NAME}"

src/apps_plugin.c

+2 -8

@@ -2387,12 +2387,6 @@ void parse_args(int argc, char **argv)

2387 }

2388 }

2389

2390 -unsigned long long sutime() {

2391 - struct timeval now;

2392 - gettimeofday(&now, NULL);

2393 - return now.tv_sec * 1000000ULL + now.tv_usec;

2394 -}

2395 -

2390 int main(int argc, char **argv)

2391 {

2392 // debug_flags = D_PROCFILE;

@@ -2456,11 +2450,11 @@ int main(int argc, char **argv)

2450 for(;1; counter++) {

2451 #ifndef PROFILING_MODE

2452 // delay until it is our time to run

2459 - while((sunow = sutime()) < sunext)

2453 + while((sunow = timems()) < sunext)

2454 usleep((useconds_t)(sunext - sunow));

2455

2456 // find the next time we need to run

2463 - while(sutime() > sunext)

2457 + while(timems() > sunext)

2458 sunext += update_every * 1000000ULL;

2459 #endif /* PROFILING_MODE */

2460

src/common.c

@@ -20,6 +20,13 @@

20 char *global_host_prefix = "";

21 int enable_ksm = 1;

23 +// time(NULL) in milliseconds

24 +unsigned long long timems(void) {

25 + struct timeval now;

26 + gettimeofday(&now, NULL);

27 + return now.tv_sec * 1000000ULL + now.tv_usec;

28 +}

29 +

30 unsigned char netdata_map_chart_names[256] = {

31 [0] = '\0', //

32 [1] = '_', //

src/common.h

@@ -37,6 +37,8 @@ extern void get_HZ(void);

38 extern pid_t gettid(void);

40 +extern unsigned long long timems(void);

41 +

42 /* fix for alpine linux */

43 #ifndef RUSAGE_THREAD

44 #ifdef RUSAGE_CHILDREN

src/main.c

+3 -1

@@ -36,7 +36,8 @@

36 #include "plugin_nfacct.h"

38 #include "main.h"

39 -#include "../config.h"

39 +

40 +extern void *cgroups_main(void *ptr);

42 int netdata_exit = 0;

@@ -84,6 +85,7 @@ struct netdata_static_thread static_threads[] = {

85 {"tc", "plugins", "tc", 1, NULL, NULL, tc_main},

86 {"idlejitter", "plugins", "idlejitter", 1, NULL, NULL, cpuidlejitter_main},

87 {"proc", "plugins", "proc", 1, NULL, NULL, proc_main},

88 + {"cgroups", "plugins", "cgroups", 1, NULL, NULL, cgroups_main},

90 #ifdef INTERNAL_PLUGIN_NFACCT

91 // nfacct requires root access

src/plugin_proc.c

+19 -35

@@ -15,12 +15,6 @@

15 #include "plugin_proc.h"

16 #include "main.h"

18 -unsigned long long sutime() {

19 - struct timeval now;

20 - gettimeofday(&now, NULL);

21 - return now.tv_sec * 1000000ULL + now.tv_usec;

22 -}

23 -

18 void *proc_main(void *ptr)

19 {

20 if(ptr) { ; }

@@ -58,7 +52,6 @@ void *proc_main(void *ptr)

52 int vdo_proc_loadavg = !config_get_boolean("plugin:proc", "/proc/loadavg", 1);

53 int vdo_sys_kernel_mm_ksm = !config_get_boolean("plugin:proc", "/sys/kernel/mm/ksm", 1);

54 int vdo_cpu_netdata = !config_get_boolean("plugin:proc", "netdata server resources", 1);

61 - int vdo_sys_fs_cgroup = !config_get_boolean("plugin:proc", "cgroups", 1);

56 // keep track of the time each module was called

57 unsigned long long sutime_proc_net_dev = 0ULL;

@@ -78,7 +71,6 @@ void *proc_main(void *ptr)

71 unsigned long long sutime_proc_softirqs = 0ULL;

72 unsigned long long sutime_proc_loadavg = 0ULL;

73 unsigned long long sutime_sys_kernel_mm_ksm = 0ULL;

81 - unsigned long long sutime_sys_fs_cgroup = 0ULL;

75 // the next time we will run - aligned properly

76 unsigned long long sunext = (time(NULL) - (time(NULL) % rrd_update_every) + rrd_update_every) * 1000000ULL;

@@ -90,11 +82,11 @@ void *proc_main(void *ptr)

82 if(unlikely(netdata_exit)) break;

84 // delay until it is our time to run

93 - while((sunow = sutime()) < sunext)

85 + while((sunow = timems()) < sunext)

86 usleep((useconds_t)(sunext - sunow));

88 // find the next time we need to run

97 - while(sutime() > sunext)

89 + while(timems() > sunext)

90 sunext += rrd_update_every * 1000000ULL;

92 if(unlikely(netdata_exit)) break;

@@ -104,7 +96,7 @@ void *proc_main(void *ptr)

96 if(!vdo_sys_kernel_mm_ksm) {

97 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_sys_kernel_mm_ksm().");

107 - sunow = sutime();

99 + sunow = timems();

100 vdo_sys_kernel_mm_ksm = do_sys_kernel_mm_ksm(rrd_update_every, (sutime_sys_kernel_mm_ksm > 0)?sunow - sutime_sys_kernel_mm_ksm:0ULL);

101 sutime_sys_kernel_mm_ksm = sunow;

102 }

@@ -112,7 +104,7 @@ void *proc_main(void *ptr)

104

105 if(!vdo_proc_loadavg) {

106 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_proc_loadavg().");

115 - sunow = sutime();

107 + sunow = timems();

108 vdo_proc_loadavg = do_proc_loadavg(rrd_update_every, (sutime_proc_loadavg > 0)?sunow - sutime_proc_loadavg:0ULL);

109 sutime_proc_loadavg = sunow;

110 }

@@ -120,7 +112,7 @@ void *proc_main(void *ptr)

112

113 if(!vdo_proc_interrupts) {

114 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_proc_interrupts().");

123 - sunow = sutime();

115 + sunow = timems();

116 vdo_proc_interrupts = do_proc_interrupts(rrd_update_every, (sutime_proc_interrupts > 0)?sunow - sutime_proc_interrupts:0ULL);

117 sutime_proc_interrupts = sunow;

118 }

@@ -128,7 +120,7 @@ void *proc_main(void *ptr)

120

121 if(!vdo_proc_softirqs) {

122 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_proc_softirqs().");

131 - sunow = sutime();

123 + sunow = timems();

124 vdo_proc_softirqs = do_proc_softirqs(rrd_update_every, (sutime_proc_softirqs > 0)?sunow - sutime_proc_softirqs:0ULL);

125 sutime_proc_softirqs = sunow;

126 }

@@ -136,7 +128,7 @@ void *proc_main(void *ptr)

128

129 if(!vdo_proc_sys_kernel_random_entropy_avail) {

130 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_proc_sys_kernel_random_entropy_avail().");

139 - sunow = sutime();

131 + sunow = timems();

132 vdo_proc_sys_kernel_random_entropy_avail = do_proc_sys_kernel_random_entropy_avail(rrd_update_every, (sutime_proc_sys_kernel_random_entropy_avail > 0)?sunow - sutime_proc_sys_kernel_random_entropy_avail:0ULL);

133 sutime_proc_sys_kernel_random_entropy_avail = sunow;

134 }

@@ -144,7 +136,7 @@ void *proc_main(void *ptr)

136

137 if(!vdo_proc_net_dev) {

138 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_proc_net_dev().");

147 - sunow = sutime();

139 + sunow = timems();

140 vdo_proc_net_dev = do_proc_net_dev(rrd_update_every, (sutime_proc_net_dev > 0)?sunow - sutime_proc_net_dev:0ULL);

141 sutime_proc_net_dev = sunow;

142 }

@@ -152,7 +144,7 @@ void *proc_main(void *ptr)

144

145 if(!vdo_proc_diskstats) {

146 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_proc_diskstats().");

155 - sunow = sutime();

147 + sunow = timems();

148 vdo_proc_diskstats = do_proc_diskstats(rrd_update_every, (sutime_proc_diskstats > 0)?sunow - sutime_proc_diskstats:0ULL);

149 sutime_proc_diskstats = sunow;

150 }

@@ -160,7 +152,7 @@ void *proc_main(void *ptr)

152

153 if(!vdo_proc_net_snmp) {

154 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_proc_net_snmp().");

163 - sunow = sutime();

155 + sunow = timems();

156 vdo_proc_net_snmp = do_proc_net_snmp(rrd_update_every, (sutime_proc_net_snmp > 0)?sunow - sutime_proc_net_snmp:0ULL);

157 sutime_proc_net_snmp = sunow;

158 }

@@ -168,7 +160,7 @@ void *proc_main(void *ptr)

160

161 if(!vdo_proc_net_snmp6) {

162 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_proc_net_snmp6().");

171 - sunow = sutime();

163 + sunow = timems();

164 vdo_proc_net_snmp6 = do_proc_net_snmp6(rrd_update_every, (sutime_proc_net_snmp6 > 0)?sunow - sutime_proc_net_snmp6:0ULL);

165 sutime_proc_net_snmp6 = sunow;

166 }

@@ -176,7 +168,7 @@ void *proc_main(void *ptr)

168

169 if(!vdo_proc_net_netstat) {

170 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_proc_net_netstat().");

179 - sunow = sutime();

171 + sunow = timems();

172 vdo_proc_net_netstat = do_proc_net_netstat(rrd_update_every, (sutime_proc_net_netstat > 0)?sunow - sutime_proc_net_netstat:0ULL);

173 sutime_proc_net_netstat = sunow;

174 }

@@ -184,7 +176,7 @@ void *proc_main(void *ptr)

176

177 if(!vdo_proc_net_stat_conntrack) {

178 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_proc_net_stat_conntrack().");

187 - sunow = sutime();

179 + sunow = timems();

180 vdo_proc_net_stat_conntrack = do_proc_net_stat_conntrack(rrd_update_every, (sutime_proc_net_stat_conntrack > 0)?sunow - sutime_proc_net_stat_conntrack:0ULL);

181 sutime_proc_net_stat_conntrack = sunow;

182 }

@@ -192,7 +184,7 @@ void *proc_main(void *ptr)

184

185 if(!vdo_proc_net_ip_vs_stats) {

186 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling vdo_proc_net_ip_vs_stats().");

195 - sunow = sutime();

187 + sunow = timems();

188 vdo_proc_net_ip_vs_stats = do_proc_net_ip_vs_stats(rrd_update_every, (sutime_proc_net_ip_vs_stats > 0)?sunow - sutime_proc_net_ip_vs_stats:0ULL);

189 sutime_proc_net_ip_vs_stats = sunow;

190 }

@@ -200,7 +192,7 @@ void *proc_main(void *ptr)

192

193 if(!vdo_proc_net_stat_synproxy) {

194 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling vdo_proc_net_stat_synproxy().");

203 - sunow = sutime();

195 + sunow = timems();

196 vdo_proc_net_stat_synproxy = do_proc_net_stat_synproxy(rrd_update_every, (sutime_proc_net_stat_synproxy > 0)?sunow - sutime_proc_net_stat_synproxy:0ULL);

197 sutime_proc_net_stat_synproxy = sunow;

198 }

@@ -208,7 +200,7 @@ void *proc_main(void *ptr)

200

201 if(!vdo_proc_stat) {

202 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_proc_stat().");

211 - sunow = sutime();

203 + sunow = timems();

204 vdo_proc_stat = do_proc_stat(rrd_update_every, (sutime_proc_stat > 0)?sunow - sutime_proc_stat:0ULL);

205 sutime_proc_stat = sunow;

206 }

@@ -216,7 +208,7 @@ void *proc_main(void *ptr)

208

209 if(!vdo_proc_meminfo) {

210 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling vdo_proc_meminfo().");

219 - sunow = sutime();

211 + sunow = timems();

212 vdo_proc_meminfo = do_proc_meminfo(rrd_update_every, (sutime_proc_meminfo > 0)?sunow - sutime_proc_meminfo:0ULL);

213 sutime_proc_meminfo = sunow;

214 }

@@ -224,7 +216,7 @@ void *proc_main(void *ptr)

216

217 if(!vdo_proc_vmstat) {

218 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling vdo_proc_vmstat().");

227 - sunow = sutime();

219 + sunow = timems();

220 vdo_proc_vmstat = do_proc_vmstat(rrd_update_every, (sutime_proc_vmstat > 0)?sunow - sutime_proc_vmstat:0ULL);

221 sutime_proc_vmstat = sunow;

222 }

@@ -232,20 +224,12 @@ void *proc_main(void *ptr)

224

225 if(!vdo_proc_net_rpc_nfsd) {

226 debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_proc_net_rpc_nfsd().");

235 - sunow = sutime();

227 + sunow = timems();

228 vdo_proc_net_rpc_nfsd = do_proc_net_rpc_nfsd(rrd_update_every, (sutime_proc_net_rpc_nfsd > 0)?sunow - sutime_proc_net_rpc_nfsd:0ULL);

229 sutime_proc_net_rpc_nfsd = sunow;

230 }

231 if(unlikely(netdata_exit)) break;

232

241 - if(!vdo_sys_fs_cgroup) {

242 - debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_sys_fs_cgroup().");

243 - sunow = sutime();

244 - vdo_sys_fs_cgroup = do_sys_fs_cgroup(rrd_update_every, (sutime_sys_fs_cgroup > 0)?sunow - sutime_sys_fs_cgroup:0ULL);

245 - sutime_sys_fs_cgroup = sunow;

246 - }

247 - if(unlikely(netdata_exit)) break;

248 -

233 // END -- the job is done

234

235 // --------------------------------------------------------------------

src/plugin_proc.h

-1

@@ -20,6 +20,5 @@ extern int do_proc_softirqs(int update_every, unsigned long long dt);

20 extern int do_sys_kernel_mm_ksm(int update_every, unsigned long long dt);

21 extern int do_proc_loadavg(int update_every, unsigned long long dt);

22 extern int do_proc_net_stat_synproxy(int update_every, unsigned long long dt);

23 -extern int do_sys_fs_cgroup(int update_every, unsigned long long dt);

24 #endif /* NETDATA_PLUGIN_PROC_H */

src/sys_fs_cgroup.c

+110 -2

@@ -14,6 +14,8 @@

14 #include "procfile.h"

15 #include "log.h"

16 #include "rrd.h"

17 +#include "main.h"

18 +#include "popen.h"

20 // ----------------------------------------------------------------------------

21 // cgroup globals

@@ -545,6 +547,33 @@ void read_all_cgroups(struct cgroup *cg) {

547 // ----------------------------------------------------------------------------

548 // add/remove/find cgroup objects

549

550 +#define CGROUP_NAME_LINE_MAX 1024

551 +

552 +void cgroup_get_name(struct cgroup *cg) {

553 + pid_t cgroup_pid;

554 + char buffer[CGROUP_NAME_LINE_MAX + 1];

555 +

556 + snprintf(buffer, CGROUP_NAME_LINE_MAX, "exec %s '%s'",

557 + config_get("plugin:cgroups", "script to get cgroup names", PLUGINS_DIR "/cgroup-name.sh"), cg->name);

558 +

559 + FILE *fp = mypopen(buffer, &cgroup_pid);

560 + if(!fp) {

561 + error("CGROUP: Cannot popen(\"%s\", \"r\").", buffer);

562 + return;

563 + }

564 + char *s = fgets(buffer, CGROUP_NAME_LINE_MAX, fp);

565 + mypclose(fp, cgroup_pid);

566 +

567 + if(s && *s && *s != '\n') {

568 + trim(s);

569 + netdata_fix_chart_name(s);

570 + free(cg->name);

571 + cg->name = strdup(s);

572 + if(!cg->name)

573 + fatal("CGROUP: Cannot allocate memory for name cgroup %s name: '%s'", cg->id, s);

574 + }

575 +}

576 +

577 struct cgroup *cgroup_add(const char *id) {

578 if(cgroup_root_count >= cgroup_root_max) {

579 info("Maximum number of cgroups reached (%d). Not adding cgroup '%s'", cgroup_root_count, id);

@@ -595,6 +624,9 @@ struct cgroup *cgroup_add(const char *id) {

624 cgroup_root_count++;

625

626 // fprintf(stderr, " > added cgroup No %d, with id '%s' (%u) and name '%s'\n", cgroup_root_count, cg->id, cg->hash, cg->name);

627 +

628 + // fix the name by calling the external script

629 + cgroup_get_name(cg);

630 }

631 else fatal("Cannot allocate memory for cgroup '%s'", id);

632

@@ -808,9 +840,9 @@ void update_cgroup_charts(int update_every) {

840 if(cg->id[0] == '\0')

841 strcpy(type, "cgroup_host");

842 else if(cg->id[0] == '/')

811 - snprintf(type, RRD_ID_LENGTH_MAX, "cgroup%s", cg->id);

843 + snprintf(type, RRD_ID_LENGTH_MAX, "cgroup_%s", cg->name);

844 else

813 - snprintf(type, RRD_ID_LENGTH_MAX, "cgroup_%s", cg->id);

845 + snprintf(type, RRD_ID_LENGTH_MAX, "cgroup_%s", cg->name);

846

847 netdata_fix_chart_id(type);

848

@@ -1061,3 +1093,79 @@ int do_sys_fs_cgroup(int update_every, unsigned long long dt) {

1093

1094 return 0;

1095 }

1096 +

1097 +void *cgroups_main(void *ptr)

1098 +{

1099 + if(ptr) { ; }

1100 +

1101 + info("CGROUP Plugin thread created with task id %d", gettid());

1102 +

1103 + if(pthread_setcanceltype(PTHREAD_CANCEL_DEFERRED, NULL) != 0)

1104 + error("Cannot set pthread cancel type to DEFERRED.");

1105 +

1106 + if(pthread_setcancelstate(PTHREAD_CANCEL_ENABLE, NULL) != 0)

1107 + error("Cannot set pthread cancel state to ENABLE.");

1108 +

1109 + struct rusage thread;

1110 +

1111 + // when ZERO, attempt to do it

1112 + int vdo_sys_fs_cgroup = 0;

1113 + int vdo_cpu_netdata = !config_get_boolean("plugin:cgroups", "netdata server resources", 1);

1114 +

1115 + // keep track of the time each module was called

1116 + unsigned long long sutime_sys_fs_cgroup = 0ULL;

1117 +

1118 + // the next time we will run - aligned properly

1119 + unsigned long long sunext = (time(NULL) - (time(NULL) % rrd_update_every) + rrd_update_every) * 1000000ULL;

1120 + unsigned long long sunow;

1121 +

1122 + RRDSET *stcpu_thread = NULL;

1123 +

1124 + for(;1;) {

1125 + if(unlikely(netdata_exit)) break;

1126 +

1127 + // delay until it is our time to run

1128 + while((sunow = timems()) < sunext)

1129 + usleep((useconds_t)(sunext - sunow));

1130 +

1131 + // find the next time we need to run

1132 + while(timems() > sunext)

1133 + sunext += rrd_update_every * 1000000ULL;

1134 +

1135 + if(unlikely(netdata_exit)) break;

1136 +

1137 + // BEGIN -- the job to be done

1138 +

1139 + if(!vdo_sys_fs_cgroup) {

1140 + debug(D_PROCNETDEV_LOOP, "PROCNETDEV: calling do_sys_fs_cgroup().");

1141 + sunow = timems();

1142 + vdo_sys_fs_cgroup = do_sys_fs_cgroup(rrd_update_every, (sutime_sys_fs_cgroup > 0)?sunow - sutime_sys_fs_cgroup:0ULL);

1143 + sutime_sys_fs_cgroup = sunow;

1144 + }

1145 + if(unlikely(netdata_exit)) break;

1146 +

1147 + // END -- the job is done

1148 +

1149 + // --------------------------------------------------------------------

1150 +

1151 + if(!vdo_cpu_netdata) {

1152 + getrusage(RUSAGE_THREAD, &thread);

1153 +

1154 + if(!stcpu_thread) stcpu_thread = rrdset_find("netdata.plugin_cgroups_cpu");

1155 + if(!stcpu_thread) {

1156 + stcpu_thread = rrdset_create("netdata", "plugin_cgroups_cpu", NULL, "proc.internal", NULL, "NetData CGroups Plugin CPU usage", "milliseconds/s", 131000, rrd_update_every, RRDSET_TYPE_STACKED);

1157 +

1158 + rrddim_add(stcpu_thread, "user", NULL, 1, 1000, RRDDIM_INCREMENTAL);

1159 + rrddim_add(stcpu_thread, "system", NULL, 1, 1000, RRDDIM_INCREMENTAL);

1160 + }

1161 + else rrdset_next(stcpu_thread);

1162 +

1163 + rrddim_set(stcpu_thread, "user" , thread.ru_utime.tv_sec * 1000000ULL + thread.ru_utime.tv_usec);

1164 + rrddim_set(stcpu_thread, "system", thread.ru_stime.tv_sec * 1000000ULL + thread.ru_stime.tv_usec);

1165 + rrdset_done(stcpu_thread);

1166 + }

1167 + }

1168 +

1169 + pthread_exit(NULL);

1170 + return NULL;

1171 +}