mirror of
https://github.com/FerretDB/documentdb.git
synced 2026-11-06 19:12:50 -05:00
Vinod Sridharan
a547931808
### Does this PR have any customer impact? Yes ### Type (Feature, Refactoring, Bugfix, DevOps, Testing, Perf, etc) Performance ### Does it involve schema level changes? (Table, Column, Index, UDF, etc level changes) No ### Are you introducing any new config? If yes, do you have tests with and without them being set? Yes ### ChangeLog (Refer [Template](../oss/CHANGELOG.md)) ### Description Currently, each round of the sproc that does create indexes does exactly 1 index build and bails. This means if a single client request queues 4 indexes and the job runs every 2 seconds, it'll take 8 seconds to build all 4 indexes. This is wasteful since we know all 4 indexes up front. Add a loop to the index build process so that multiple indexes can be built without waiting the 2 seconds. As part of this, there were a few places allocating in memory contexts that are unstable across executions. Fix this. Additionally there was a loop that could cause index builds to starve for ~20 mins if it picked the wrong index. Fix this loop too. Finally, Noticed that the version cache does not cache properly if the Major version is 0. Fix this (which can cause perf regressions in DocumentDB).