Concerns about issuing drop measurement on large database


I have a large telegraf database, with 1 year worth of data for ~300 systems, total 950G. Daily shards, 1 min resolution for the most part.

Early on I decided to use the interrupts plugin without any filtering or aggregation (facepalm). Today, on a 2.2GB shard, 600MB’s of those are for the interrupts measurement (and interrupts is the only plugin being collected at 5 mins interval). And that is absurd. It’s a lot of systems and with many cpu’s.

I just want to clean it up, and use a simpler interrupts setup, where I just save the total for all cpu’s for each IRQ, which I’m sure will save space. I don’t mind losing the interrupts measurement, but I do mind about all other measurements on that database.

However, I’m concerned about the effect of just issuing “drop measurement interrupts” at this scale, as I’m not sure of the mechanics involved in that operation.
Will InfluxDB be able to handle this without issue, or will try to put a lot of stuff into RAM and eventually crash? Or take forever reshuffling data in the tsm files?

Any hints on this?


Ok then…
Thanks for the feedback!

Lol, maybe simply just give it a shot. Or in order to test it, create another influxdb instance in new container copy your db over there and see how it goes.
Nonetheless my experience is that DROP MEASUREMENT takes less of hw resources than for example DELETE FROM