This is an example for https://issues.apache.org/jira/browse/MESOS-2713.
Docker stats and collectd-docker report correct values from cgroups.
Mesos slave with /monitor/statistics.json endpoint reports wrong data.
This happens because CPU usage is calculated on a tree and recalculated
each time. When process dies, all of it's usage is gone as well.
Three consecutive measurements:
[
{
"statistics": {
"timestamp": 1431340786.14526,
"mem_rss_bytes": 671744,
"mem_limit_bytes": 570425344,
"cpus_user_time_secs": 0.72,
"cpus_system_time_secs": 0.06,
"cpus_limit": 1.1
},
"source": "wtf_wrong-stats.ce16b7c8-f7c9-11e4-a977-56847afe9799",
"framework_id": "20150126-100650-3909200064-5050-1-0007",
"executor_name": "Command Executor (Task: wtf_wrong-stats.ce16b7c8-f7c9-11e4-a977-56847afe9799) (Command: NO EXECUTABLE)",
"executor_id": "wtf_wrong-stats.ce16b7c8-f7c9-11e4-a977-56847afe9799"
}
][
{
"statistics": {
"timestamp": 1431340786.86123,
"mem_rss_bytes": 671744,
"mem_limit_bytes": 570425344,
"cpus_user_time_secs": 1.43,
"cpus_system_time_secs": 0.07,
"cpus_limit": 1.1
},
"source": "wtf_wrong-stats.ce16b7c8-f7c9-11e4-a977-56847afe9799",
"framework_id": "20150126-100650-3909200064-5050-1-0007",
"executor_name": "Command Executor (Task: wtf_wrong-stats.ce16b7c8-f7c9-11e4-a977-56847afe9799) (Command: NO EXECUTABLE)",
"executor_id": "wtf_wrong-stats.ce16b7c8-f7c9-11e4-a977-56847afe9799"
}
][
{
"statistics": {
"timestamp": 1431340787.79716,
"mem_rss_bytes": 671744,
"mem_limit_bytes": 570425344,
"cpus_user_time_secs": 0,
"cpus_system_time_secs": 0.02,
"cpus_limit": 1.1
},
"source": "wtf_wrong-stats.ce16b7c8-f7c9-11e4-a977-56847afe9799",
"framework_id": "20150126-100650-3909200064-5050-1-0007",
"executor_name": "Command Executor (Task: wtf_wrong-stats.ce16b7c8-f7c9-11e4-a977-56847afe9799) (Command: NO EXECUTABLE)",
"executor_id": "wtf_wrong-stats.ce16b7c8-f7c9-11e4-a977-56847afe9799"
}
]