We have a application which stores the timeseries data in influx the rate of ingest to influx db is around 100/sec. We are noticing that influxd process CPU spikes to 500% for sometime and then comes back to steady state in couple of minutes. What could be the reason for this spike? We don’t have any get query at this point, everything is create to influx. Is there any config we have to tweak the process given to influx we have enough CPU on the VM
just two things come to mind:
- continuous queries (frequency defined in the definition itself, gets logged)
- data compaction (frequency somehow depends on the config, gets logged)
If that’s the case the issue should present at a fixed frequency.
yes we do have Continuous query running every 15 mins and we notice spike during this time.If it’s continous query, how to go about fixing it?
if it’s due to CQs running there is no magic solution, look at the queries and see if they can be:
- improved → whatever it may mean… it depends on too many factors (logic, complexity, amount of data)
- change their frequency → some overlapping will occur anyway, but not every single time they run (maybe once 2-3-4 runs there will be overlapping, it all depends on frequency)
- deleted → if un-needed/obsolete
