Open beevee opened 5 years ago
What i understood:
NODATA -> OK -> NODATA
metrics in redis, which is ofcourse false info. NODATA
metric which in turn will give number of active metrics (total metrics - counter).counter > curr_active_metrics
.@hv7214 I imagined something like this:
When a user sets NODATA == DEL for a trigger, allow user to specify the desired amount of active (i.e. not deleted) metrics. Notify when there are more or less metrics than expected.
@beeve i want some more details : Notif has to be sent once per crossing the marked metric level, or after every 24 hrs if metric count is less or more ?
@hv7214 I think once.
We are stepping into the new world of orchestrated clouds. Long-living pretty-named hosts are gone and replaced by short-lived containers with some random characters as a host name.
Yesterday we had a pattern
DevOps.*.cpu.percent
that matched:DevOps.host1.cpu.percent
DevOps.host2.cpu.percent
DevOps.host3.cpu.percent
Today it matches a lot more metrics:
DevOps.container-93EE63.cpu.percent
DevOps.container-CD91FF.cpu.percent
DevOps.container-177761.cpu.percent
DevOps.container-35DE27.cpu.percent
This poses two problems with Moira:
A solution for (1): let user set a desired count of active metrics in trigger. Notify user if there are less active metrics than expected. Metric is considered active if there are datapoints inside the TTL window (same criteria as for NODATA state).