Recently we have upgraded our Couchbase cluster from 4.5.1 CE to 6.0 CE.
We have 3 nodes with all services (index,data,query,search actually we are not using search seems it came automatically)
Most time CPU used by indexer service uses 100% of CPU in one node. Why it is using that much CPU. Also why it is happening in only one node. Another two is perfectly fine. Anyone can help me to find the cause? Thank you in advance. Top command attached
After a few investigations, Found that the CPU is increasing during the index fragmentation interval.
Is it due to circular write mode? What will happen if I change to Append-only mode.
Also currently the index in the same node consumes more disk than other nodes. Is it due to any issue with the Fragmentation process? Please share your thoughts even if it is not a solution.
One thing I suspect is,
While we configuring the cluster, the Index storage mode is Append-only.
After adding two more nodes(total 3 nodes), we updated the storage mode to Circular write mode from Node 1’s console.
As per documentation, changing storage mode should be done by removing existing index nodes. But as we changed without removing other nodes, the change is reflected in Node 1 only.
That can be the reason why Node 1 only showing strange behaviour. This can be a reason?
We have figured out the cause and fixed the issue. Posting here if it would be helpful to others.
We were using old Couchbase Java client in our server application even after upgrade. There was no compatibility issue with the latest CB 6.0 CE.
After upgrading the Couchbase Java client, the cluster became normal.
1 Like