Sergii Kreitor

Update. We were checking pod consumption during the day and recently - over the night. As you can see consumption slowly grows. Pod was running for 22 hours without restarts...

If we spinup the same container without datasources subscribed - memory consumption does not increase. Deviations are mostly due to OPA memory usage. And again, no traffic on both pods.

@obsd the only thing we can think that can cause this problem is fetching/updating or backing up mechanism. Is there any progress on this? It is kinda urgent for us...

Hi @gemanor @sroopsai . I'll try to make a reproducible example this week. To give more info - we have opal server in 2 replicas and kafka as a backbone....

Hi @gemanor, sorry, haven't got a chance to try to reproduce this. was focused last week more with this https://github.com/permitio/opal/discussions/671 as it affects us more. I'll let you know the...

Yeah, we actually worked with OPAL for a while in a previous setup. However, some performance and optimization issues pushed us to move fully to OPA with a set of...

Yes, we did. You can take a look at the profiling results attached at the bottom of the PR description. Object/struct pools noticeably reduced the number of allocations, which improved...

There are 3 files original.txt, optimized.txt and comparsion.txt. The last one is generated by `benchstat`. ``` benchstat original.txt optimized.txt ```

There is a section in description that explains exactly the thing that you've asked ``` Performance Validation The PR includes detailed benchstat comparison (benchstat original.txt optimized_fixed.txt) showing: - Consistent improvements...

Hi @orweis , I'll add more details. It doesn't throw the error if you don't send the updates (policy/data) to opal server (via REST API call or via direct broker...