Details
-
Improvement
-
Resolution: Fixed
-
Critical
-
None
-
Security Level: Public
Description
The cluster manager works by 'electing' a single orchestrator node to perform certain tasks such as rebalance orchestration or failover.
This means that you would want to direct most administrative REST calls to this node, as it avoids extra hops from nodes proxying to the orchestrator.
Additionally this node's logs will contain more information about rebalances, failovers etc than any of the other nodes so you'll often want to collect these logs first in a failure scenario.
For that reason we should add a stable API to identify the orchestrator node in a given cluster that doesn't rely on diag/eval (as that's locked behind localhost now anyway).
Attachments
Issue Links
For Gerrit Dashboard: MB-31649 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
138007,3 | MB-31649: terseClusterInfo returns limited information by default | mad-hatter | ns_server | Status: MERGED | +2 | +1 |
138096,3 | MB-31649: Maintain same terseClusterInfo in UI | mad-hatter | ns_server | Status: MERGED | +2 | +1 |
138265,1 | Merge remote-tracking branch 'couchbase/mad-hatter' | master | ns_server | Status: MERGED | +2 | +1 |