I talked to people about this at PromCon recently; the consensus was that Prometheus does not expect that Labels.Hash() is constant across runs of Prometheus.
Possible issues: * The 'hashmod' function in relabelling: this uses a MD5 hash, separate from Labels.Hash(). * Alert.String() calls Labels.Hash(). I could not find where this was used. I created https://github.com/prometheus/prometheus/pull/11635 to add a comment warning it may change. Bryan On Wednesday, 9 November 2022 at 15:11:15 UTC Bryan Boreham wrote: > As part of #10991 I took advantage of the new data structure to make > Hash() run a lot faster; however the new version gives different answers to > the old version. From my reading of the Prometheus code, this is OK because > hash values are never persisted. > > Later, I became aware that query-sharding in Mimir used this Hash function > in persisted values. That is on Mimir (i.e. me) to fix. > > My question to the group is: are there likely to be other places where it > was assumed this value will never change? I can change the labels.Labels > API to have a new FastHash() or similar which is used in > performance-sensitive places, and reinstate the old slow one. > > Either way I would add to the docs whether Hash() should be expected to > change. > > Bryan > > -- You received this message because you are subscribed to the Google Groups "Prometheus Developers" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/prometheus-developers/fbd9b34f-c6c6-4877-98ad-e517a8218acbn%40googlegroups.com.
