Details
-
Bug
-
Resolution: Won't Fix
-
Major
-
1.8.1
-
Security Level: Public
-
None
-
centos 5.4 64 bit
Description
Install couchbase server 1.8.1-840 on 10 centos 5.x servers 64bit
Create a 10 nodes cluster with 6.5 GB default bucket.
Load 30 million items with 256 bytes/item and 10 million items with 512 bytes/item
Cluster goes in to dgm with 78.8 % resident ratio
Mutate 40 million items to create cache miss and disk fetch. Maintain the load around 15K to 17K ops.
Add a node (54) to cluster and rebalance
Add another node (55) to cluster and rebalance
Remove node 54 and failover node 55. Then rebalance.
In all cases above, the drain rate is very slow (around 150 to 300 items per second)
View drain rate at each node, drain rate is zero at some nodes
Disk write queue steady increase