Low frequency samples ingestion #7666
Unanswered
bodulos
asked this question in
Help and support
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi,
I'm migrating my current prometheus long-term storage setup (promscale+timescaleDB) to Mimir and have some challenges with ingesting low frequency samples. Because of the nature of the datasources we need to collect some offline generated timeseries data (jobs, timeseries file dumps etc.) with sample frequency of 1 hour, 1 day etc. Those samples have their embedded timestamps which we need to honor.
This is not an issue for current setup, promscale accepts older timestamps in samples and handles older chunk decompression before ingestion.
Because of the default restriction in Mimir samples older than 1h cannot be ingested. I am aware that this could be tuned but not sure this is the right approach.
From what I have read, ingesting samples more than 24 hours old samples falls in domain of data backfill. Idea is to use the backfill method to ingest older samples - create prometheus format chunks from timeseries data and upload them with mimirtool.
Hourly samples are border case. Changing the ingestion window from 1h to 2h in would solve problem for them. But could be handled in the same way as daily samples, with backfill.
Before I proceed in this direction I have few questions to confirm it:
Any thoughts on the subject would be highly appreciated.
Beta Was this translation helpful? Give feedback.
All reactions