Raúl Marín

> May be related to https://github.com/ClickHouse/ClickHouse/pull/35956, cc: @Algunenano It could be, but it's pretty hard to say without more logs.

To be this looks like a race condition. It seems possible if 2 threads call `DNSResolver::updateCache` since, after my changes, we might be inserting and dropping elements from the `std::unordered_map`...

I've looked into this and although there should be better mutex handling to make sure `DNSResolver::updateCache` isn't running multiple times at the same time (since that has the change of...

I don't know if it's related or pure luck, but one thing that caught my attention is that it seems to always crash on `updateAddress` and never on `DNSResolver::updateHost`, although...

Looking at the worse performance drop (22.2 -> 22.3) it was introduced in: ``` fe4534d4646e51e40ac04fec83e4a321c4ce19ea is the first bad commit commit fe4534d4646e51e40ac04fec83e4a321c4ce19ea Author: Amos Bird Date: Mon Mar 7 21:51:34...

> Must be fixed by https://github.com/ClickHouse/ClickHouse/pull/31796. For reference I tested the branch and got similar performance as in master: master (5ebf230dbbffb9b087047b31b03a9c5228eb2ae2): ``` localhost:9000, queries 100, QPS: 8.149, RPS: 0.000, MiB/s:...

> this branch introduce new analyzer and planner, it is not enabled by default you need to specify use_analyzer = 1. I see, that's for the pointers. I'll keep an...

https://s3.amazonaws.com/clickhouse-test-reports/40002/c8f8ceea8d5694eaf5ad4df41297a55e932d6fcd/stress_test__undefined_.html ``` SELECT status, count() AS all FROM test ARRAY JOIN nested as nestedJoined WHERE (status IN ( SELECT status FROM test ARRAY JOIN nested as nestedJoined GROUP BY status...

I did this a while back (months before implementing wagyu inside Postgis):  This was comparing our tiler serving MVTs using St_AsMVT or using node-mapnik. The main difference I've noticed...

You might want to update wagyu too. 0.5.0 should be faster in some cases and more robust.