Details
-
Bug
-
Resolution: Duplicate
-
Critical
-
3.0
-
Security Level: Public
-
None
-
- Centos 6.4
- Cluster : http://10.6.2.163:8091
- Configuration : VMs- 1 SSD 2 Partitions /data , /index
- Build : 3.0.0-858-rel
- 2 Buckets, Per bucket 1 DdocX 2 Views,
- View & Data compaction set at 30%
-
Untriaged
-
Unknown
Description
1. Setup a 7 Node cluster, with above setup
2. Load items ~ 80M on both the buckets
3. Seeing very uneven view compaction across the nodes for both the buckets.
- View compaction does not kick on some of the nodes, while others behave as expected.
- View Disk size grows enormously on the saslbucket
4. End of the phase, user-triggered compaction is the way to even out the view-disk usage
5. Likewise behaviour for data-compaction as well, it exceeds the 30% threshold level and does not compact across some nodes.
Attached screenshots of nodes showing high fragmentation - "couch_disk_size"
- Seeing this on multiple runs of the same test
Attachments
Issue Links
- duplicates
-
MB-10273 View compaction should be triggered predictably, especially with parallel compaction enabled
- Closed