Open jnpkrn opened 4 years ago
Due to time constraints and other priorities, we haven't been able to implement these planned higher level booth commands which would be aware of the whole booth formation and which would be able to operate at a cluster level and booth formation level. For the users to be able to set booth at least somehow, only the node-level commands have been implemented. Those, even if not as comfortable as one may desire, provide a complete set of operations needed for configuring and managing booth.
The plans for the more comfortable booth management are still there. Unfortunately given current developer resources and backlog, it is a "blue-sky" project for now.
In case you happen to run said command in case you didn't pick the right, "currently representative" node, you get to see an unexpected outcome:
Well, of course, you cannot, since pcs invoked you at the wrong node (lazily, just the local one right away).
What should pcs have done instead:
realize whether the home node is part of the cluster with multisite clustering configured
if so, it shall deduce where the local cluster representative is currently running within cluster, and route the request over there[*]
if not, currentl behaviour is likely fine, expect to report from local arbitrator POV
[*] regarding "routing the request (soliciting info), there are 2 possiblities:
A. utilize presumed pcs intra-node communication infrastructure, route the request to the target node using pcs-native means, than route the response back the same way
booth list
andbooth status
B. utilize "distributed/non-local connectivity" of booth, allowing for
booth status
command to return non-success and replacing thebooth list -c booth
command withbooth list -s <booth virtual IP>
booth list
side of the report, hence A. is strongly recommended (when applied within cluster, you want to get the same information back regardless if where the cluster-local booth representative is running)