Details
-
Bug
-
Resolution: Fixed
-
Critical
-
7.6.0
-
Untriaged
-
0
-
Unknown
Description
This was found during code inspection :
https://github.com/couchbase/plasma/blob/master/recovery.go#L1746
Justification :
During rebalance recovery at destination node, we can update the memory quota based on checkpoint data of a shard on source node. As the source and destination node can have different set of instances, the memory quota adjusted by memtuner may not be applicable to current node. Although mem-tuner can re-adjust quota based on RSS and free system memory, this can cause some jittery in performance during rebalance or slowness in overall rebalance duration if the updated memQuota is too low.
The fix should be simple and do not expect to cause regression.
Attachments
For Gerrit Dashboard: MB-60529 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
203558,4 | MB-60529: Log pageRemove key in plasma dump | unstable | plasma | Status: MERGED | +2 | +1 |
204298,4 | MB-60529: Do not update memQuota during recovery of restored shards | unstable | plasma | Status: MERGED | +2 | +1 |
204505,2 | MB-60529: Fix return on error in rpc handler | unstable | plasma | Status: MERGED | +2 | +1 |
204532,2 | MB-60529: Lower log level for rpc client | unstable | indexing | Status: MERGED | +2 | +1 |
204606,1 | Merge bug fix for MB-60529 from unstable to main | master | plasma | Status: MERGED | +2 | +1 |
204663,4 | Merging fixes for MB-60509,MB-60338,MB-60534,MB-60529 | master | indexing | Status: MERGED | +2 | +1 |
205213,1 | MB-60529: Lower log level for rpc client | unstable | indexing | Status: MERGED | +2 | +1 |
205221,2 | Merging fixes for MB-60338, MB-60529, MB-60534, MB-60509 and MB-60580 | master | indexing | Status: MERGED | +2 | +1 |