hi guys,
# spicedb
s
hi guys, I'm sorry to bring up this question again, but I still can't figure out what's going on with the datastore metrics. particularly, QueryRelationships RPS are a couple times higher than pgxpool acquire rate. any help is appreciated 🙂
v
hey @sashayakovtseva_46690 since you've been showing interest in datastore metrics, this may be relevant: I've opened https://github.com/authzed/spicedb/pull/1669 The gist is that the proxy used to measure datastore access is installed in a place that does not represent real access to the actual underlying datastore. The caching and deduplication proxies are instaled in between. This means for example that you may have counted 1
QueryRelationship
API call, but that actually may not have happened under the hood becayse it was either cached or deduplicated.
also please note that there is no guarantee 1 Datastore API call = 1 SQL query. But in most cases it should.
s
thanks! I've also found out that 1) you've duplicated metrics with double observable proxy (https://discord.com/channels/844600078504951838/1177657599747956918/1179072705274859580) and 2) also measured metrics multiple times in close (https://github.com/authzed/spicedb/issues/1664) I'm looking forward for any further improvements, I think dtastore metrics are really important b/c they correlate with SpiceDB performance
v
@sashayakovtseva_46690 duplicates are fixed in 1.28.0, indeed
1.28.0 just released yesterday
https://github.com/authzed/spicedb/issues/1664 didn't make it into 1.28.0 and will be part of the next release
s
got it, I actually switched to main branch builds for further investigation thanks again for reaching out and letting me know 🙂
v
no worries! there is also a nightly build of SpiceDB so you don't have to build yourself: https://hub.docker.com/r/authzed/spicedb-git