add log to identify request cache hit and cache miss

[mengmengy]

Signed-off-by: Mengmeng Yang [email protected]

What this PR does: Add few more logs to cache results file so that it's easier to figure out a request is skipped cache, cache hit, cache miss when a request split to multiple ones.

Checklist

• [ ] Tests updated
• [ ] Documentation added
• [ ] CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]

[alanprot]

We dont have metrics for it already?

[mengmengy]

We dont have metrics for it already?

I think it logs few counters like cortex_cache_hits, cortex_cache_fetched_keys (like below), but from these, we cannot tell which split request got cache hit or cache miss. Please advise if I missed anything.

// TYPE cortex_cache_fetched_keys counter cortex_cache_fetched_keys{name="frontend.redis"} 3

// TYPE cortex_cache_hits counter cortex_cache_hits{name="frontend.redis"} 1

[alvinlin123]

I am not sure if we should do info logging on the call path that may be high TPS. May have impact on latency.

[mengmengy]

I am not sure if we should do info logging on the call path that may be high TPS. May have impact on latency.

Understood, not all logs will be executed for each split request, for each split request, only 1 log is added. Add such log will be useful when we try to figure out issues like why request is being slow etc. Or do you suggest other better ways we can figure out this I'm not aware?

[alvinlin123]

Maybe have it to be debug log instead of info, or maybe like @alanprot suggested, use metrics instead of logs.

[mengmengy]

Maybe have it to be debug log instead of info, or maybe like @alanprot suggested, use metrics instead of logs.

Metrics may not suitable for this case since it will be good to know which split request get cache hit/miss... I updated log level from info to debug.

[mengmengy]

@alanprot @alvinlin123 could you please take a review about the updated PR? Thanks