Community

banshee is under active development: collectors will keep changing, and a version you publish will likely need adjustments. For stable use, wait for version 1.0.0.

Examples

banshee/examples-collection/examples/watches/

increase

Example: the watch increase on a counter over a window; out-of-memory kills by SSH.

  • example
  • watch
  • increase
  • counter
  • memory
  • ssh

by banshee1 installation1 version

v1Latest

Published on October 27, 2026. First version in this collection.

Watch: increase

watch.increase raises when a counter, a number that only grows, grew too much over a window: here, the processes the kernel killed for lack of memory.

How to read it

  • The output is the raw counter, oom_kill of /proc/vmstat, since the start of the machine. The collector does not compute any difference: the watch does.
  • 'window' => '1 d': the increase is counted over the last day. 'warning' => 0: any kill in the day is a warning, since the increase must be strictly above the threshold; 'critical' => 2: three kills or more.
  • A value smaller than the previous one, after a restart, makes the window start again from it.

To adapt it

  • Any counter fits: carrier changes of an interface (/sys/class/net/eth0/carrier_changes), errors of a disk, restarts of a service.
  • oom_kill exists since Linux 4.13; on an older kernel, grep finds nothing and the collector stops.