Closed todvora closed 10 months ago
The same applies if you click "resume startup" without provisioning available data node(s).
In this case, the problem is even worse because the only way to resolve it afterwards is to set the enable_preflight_web
flag which has to be removed after the first startup again.
I think we should add the SKIPPED state. This could be used to determine if the preflight UI should be shown again (SKIPPED && datanodes available) or if the default connection should be used (SKIPPED && no datanodes available)
Alternatively, I would suggest making the text in the confirmation dialog more explicit:
"Are you sure you want to resume startup without a configured Graylog datanode? This requires pointing elasticsearch_hosts
in graylog.conf
to a running Opensearch or Elasticsearch instance."
Also, I would change the comments in misc/graylog.conf
to not specify a default anymore but explain the new mechanism.
Let me know what you would prefer. I will then implement that solution. I am in favor of the first solution
Probably, a combination of both of the above would be optimal
Could we create a backport to 5.2 for this, please?
If the user opens the preflight configuration for the first time and decides to "resume startup" without configuring any datanodes, the app, due to IndexerDiscoveryProvider, triggers IllegalStateException.
Expected Behavior
Before datanode and preflight, the default behaviour was to use whatever listens on
http://127.0.0.1:9200
if the user does not configure anything else.Current Behavior
We don't discover any datanodes but we see that the user finished the preflight (
PreflightConfigResult.FINISHED
), so we asume the datanode path and fail immediately.Possible Solution
Steps to Reproduce (for bugs)
Context
Your Environment