HCD-92 Upgrade Jamm version to 0.4.0 #1631
Conversation
Backport of 600f4d9 from the main-5.0 branch.
|
We will need to test this on CNDB, and we have to update CNDB scripts to load jamm and |
|
I missed some errors in the test code, switching to draft while I fix them. |
|
✔️ Build ds-cassandra-pr-gate/PR-1631 approved by ButlerApproved by Butler |
There was a problem hiding this comment.
This is why MemtableSizeHeapBuffersTest was failing.
Until now, ObjectSizes.meter was built with omitSharedBufferOverhead(). But with the changes in this PR, it isn't, so we need a dedicated method (copied from 49e0c611 on the main-5.0 branch).
FWIW, adding logs in SkipListmemtable#estimateRowOverhead() helped pinpoint the issue, the estimated sizes between main and this PR were wildly different.
There was a problem hiding this comment.
FWIW, adding logs in SkipListmemtable#estimateRowOverhead() helped pinpoint the issue, the estimated sizes between main and this PR were wildly different.
For debugging purposes, you can also use .printVisitedTree() to get a printout of all the objects
and their calculated size on heap to debug jamm related issues. See:
README.md on github
|
Sorry I need to close this -- by pushing my branches to a fork I accidentally polluted a bunch of apache PRs. I will reopen next week from a branch in this repo. |
|
Reopened as #1649. |
4 participants
Backport of 600f4d9 from the main-5.0 branch.