During a server update (actually a docker update), we had to restart couchbase and it took at least 15 minutes for the bucket (we only have one) to be available. It was in pending mode.
15 minutes unavailability is not acceptable in production environment. But I suppose there is something we are doing the wrong way.
Is it usual for a bucket to stay so long in pending mode ? What can be done to fix this ?
We are using 4.5.1-2844 Community Edition (build-2844)
Thanks for your help.
Can you post some details about your cluster and bucket? Like number of documents/views/indexes and the total size of the bucket contents? Also useful would be the number of nodes you use and your replication settings (i.e. how many replica copies if any?).
Your assumptions are sound – this is unusual in a typical Couchbase setup unless there are unusually large amounts of data and re-indexing going on.
Thanks for your answer.
We have 1 node, 1 bucket, 26 views, no indexes (visible in the admin console), no replication, 3,027,683 documents. Backup size is around 500MB.
I suspect that our config is far from optimal and good practices. I would be more than happy to get your insights what should be done with highest priority given that we are no expert in CB and don’t have much time to become one.
It does look like you have graduated to needing a multi-node cluster, but there’s nothing alarming in your usage, beyond the sheer number of documents. Try this guide: https://developer.couchbase.com/documentation/server/4.6/install/sizing-general.html
I should also mention (though I understand that in your use case it was necessary) that “restarting” Couchbase is generally not done in a production environment, since Couchbase Server can easily rebalance your activity across multiple server nodes (really a minimum of four assuming you want replication) even if you add and remove nodes. So, for instance, this is how we recommend you upgrade your version of Couchbase Server: 1 node at a time, live.