Details
-
Improvement
-
Resolution: Unresolved
-
Major
-
6.0.0, 6.6.0, 6.5.0, Cheshire-Cat
-
1
Description
The fact that nodes get more-or-less completely wiped upon removal from a cluster trips folks up from time-to-time. Quite frequently these nodes get removed for reasons of periodic maintenance (OS security patches) or upgrade of Couchbase Server and the node is destined to be re-added to the same cluster in short order. The fact that the customer needs to largely reconfigure the node can trip people up.
A good example of this is the case when users are using their own certificates (i.e. not the auto generated ones.) When the node gets removed these get wiped. If the user forgets to re-upload the CA and node cert prior to re-adding the node via the HTTPS port (which is the default) they'll hit an unknown CA error. Uploading certs is a tedious enough task that this is a legitimate annoyance.
It would be nice to support a notion of "node removal for maintenance" or maybe more generically "temporary node removal" where the node is removed but it will be added back to the same cluster later.
In these cases the node could retain the following information:
- its name
- cluster CA and node certificate
- services that it is running
- alternate address configuration
- stats
And perhaps other things. I'd stop short of having it retain data so that the node could be added back via delta node recovery - that feels like a different feature.