Details
-
Bug
-
Resolution: Won't Fix
-
Critical
-
3.0.1
-
Security Level: Public
-
Untriaged
-
Windows 64-bit
-
Yes
-
Mar 9 - Mar 27
Description
Hardware:24 vCPU, 64 GB RAM, HDD
Scenario:
2<->2 nodes bidirectional XDCR, 2 buckets x 25M x 2KB
DGM:resident ratio(40%)
Bucket RAM quota per cluster: 40GB
no front end workload running during XDCR phase i.e XDCR is initiated after load on both clusters.
XDCR replication on 2.5.1 cluster completes. However on 3.0.1 there are hard OOMs on one of the clusters and XDC replication is stuck. I increased the Bucket RAM quota by 10% to 44GB but still see the same issue. From the mem_used on the 2.5.1 run, the mem_used on 3.0.1 is 20% higher, yet XDCR is stuck.
links:
3.0.1 logs are at:
cluster2 logs
https://s3.amazonaws.com/cb-customers/performancetestsxdcr2x2sourcecluster/performance/collectinfo-2014-10-02T212416-ns_1%40172.23.96.26.zip
https://s3.amazonaws.com/cb-customers/performancetestsxdcr2x2sourcecluster/performance/collectinfo-2014-10-02T212416-ns_1%40172.23.96.25.zip
cluster1 logs
https://s3.amazonaws.com/cb-customers/performancetests/performance/collectinfo-2014-10-02T212335-ns_1%40172.23.96.27.zip
https://s3.amazonaws.com/cb-customers/performancetests/performance/collectinfo-2014-10-02T212335-ns_1%40172.23.96.28.zip