Details
-
Improvement
-
Resolution: Unresolved
-
Major
-
5.0.0
Description
As part of query monitoring, system keyspaces should throw warnings to the effect that they are skipping nodes that are known to be down, but this can't be done because n1ql clustering only reports active nodes.
Warnings get currently issued only for those nodes that go down since ns_server last tracked their status.
This MB is to track improvements to clustering to the effect that all nodes are tracked, with status.
This was originally part of MB-22709.
The erroneous behaviour described in MB-22709 has been fixed, so we might just as well treat this as a separate improvement.
Attachments
Issue Links
- relates to
-
MB-30992 Fetch on completed requests very slow
- Closed
Gerrit Reviews
For Gerrit Dashboard: MB-24229 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
98649,3 | MB-30992 clustering improvement cache known query nodes for performance | master | query | Status: MERGED | +2 | +1 |
105594,3 | MB-33171 avoid connection leak from go-couchbase to ns_server | master | query | Status: MERGED | +2 | +1 |