Details
-
Bug
-
Resolution: Duplicate
-
Major
-
7.1.3
-
Untriaged
-
0
-
Unknown
Description
A test on AWS usingĀ instanceType: "a1.2xlarge" ran into the OOM issue. The reclaim_pending reaches 2GB before the OOM issue.
I am adding the tar zipped file that has the cbcollect (indexer.zip), metrics from cloudwatch, a set of goroutine dumps, iostats and heapdump profiles when the free memory is low.
The test has one bucket, one index
inserts 300M docs, creates an index on a 128 byte key (field +docID is 128 bytes). Then 150M keys are updated. The index keys are random. During the incremental build the OOM is seen. After the OOM, the indexer restarts and the update completes.
Attachments
Issue Links
For Gerrit Dashboard: MB-55188 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
185499,2 | MB-55188: Stop eviction when heap limit is set | neo | plasma | Status: NEW | 0 | 0 |