Confusion between cluster and node-level RAM quotas
Description
Components
Affects versions
Fix versions
Labels
Environment
Link to Log File, atop/blg, CBCollectInfo, Core dump
Release Notes Description
Attachments
- 11 May 2017, 06:04 PM
Activity
Rob Ashcom January 10, 2020 at 7:35 PM
The core confusion here that I can address is whether the user is tweaking settings for the node or the cluster level. This is always a bit tricky with configuration settings in Couchbase, but the UI can do better.
Mike Wiederhold May 23, 2017 at 5:51 AM
Moving to the feature backlog. It might be a good idea to see if we can make these pages and setting memory quotas a little bit more intuitive.
Dave Finlay May 12, 2017 at 5:56 PM
Perhaps, given this info, you should state what you believe the bug to be? 🙂
Aruna Piravi May 12, 2017 at 5:55 PM
Thanks Dave!
Dave Finlay May 12, 2017 at 5:53 PM
Note that if FTS and Indexing are not running on a given node, all memory is available to the data service. You'll notice this in cluster setup; if you disable FTS, say, the memory used drops by the FTS memory quota number. We show the numbers on the cluster settings page - but they only get added on nodes that are running all the services.
Build
5.0.0-2837
1. Install CB server on a single node with 'data' service.
2. Check the perServer RAM quota. I see fts + GSI taking almost 1GB RAM.
It's OK if I have these services present in the cluster but I don't, in which case, I should atleast be allowed to set the ram quotas for FTS and GSI to 0MB, but an error is thrown if I try to.
These are called perServer RAM quota but unfortunately it is set at the cluster level.
Pls see attached screenshots. I did a quick search in Jira about this issue and didn't get relevant hits. Please close as duplicate if this has been filed already.