Metrics

Inspeqtor collects two types of metrics:

• overall host-based metrics for the machine
• per-process metrics for each service being monitored

Metrics are named like memory:rss where memory is the family and rss is the metric within that family.

Metrics

Host-wide

• load:1
• load:5
• load:15
• swap
• cpu:user
• cpu:system
• cpu:steal
• cpu:iowait
• disk:$MOUNTPOINT

disk is a special "dynamic" metric, one for each entry in df with a real device:

$ df
Filesystem     1K-blocks    Used Available Use% Mounted on
/dev/xvda       20131844 3420500  16506776  18% /
none                   4       0         4   0% /sys/fs/cgroup
devtmpfs         1019296       4   1019292   1% /dev
none              204120     176    203944   1% /run
/dev/xvdc        8029632 2296928   5405028  30% /old

The first and the last entries are real filesystems so Inspeqtor would have two disk metrics: disk:/ and disk:/old.

Per-process

• memory:rss
• memory:total_rss
• cpu:user
• cpu:system
• cpu:total_user
• cpu:total_system

The total_* metrics include all child processes. This is useful for forking servers like Apache or Unicorn where the parent coordinator process doesn't do much and you want to monitor the child processes for memory leaks/bloat or unexpected CPU spikes.

Have ideas for other metrics to track? Open an issue with your use case.

Daemon-Specific

See Daemon-Specific Metrics for the metrics you can monitor for each supported type of daemon.