Open milanage opened 3 years ago
It seems the timeout is 1 minute when grabbing Kibana diag.
23:57:41.776 [main] INFO com.elastic.support.diagnostics.commands.CheckKibanaVersion - Getting Kibana Version.
23:58:41.875 [main] ERROR com.elastic.support.rest.RestClient - Unexpected Execution Error
Not very sure if it's a pure timeout issue yet, but given it's hard to tweak timeout value as of now, due to it's not exposed as parameter. Could we please first expose this option to external and see if by simply tweaking timeout value can solve the issue?
Or alternatively maybe we can use 5 minutes by default, but make it tuneable + default a bit longer may be better, based on the situation.
These are timeouts that are defaulted in the stack diagnostics tool not in eck-diagnostics
https://github.com/elastic/support-diagnostics/blob/bad8fe76f2d2be716c14ffc5455f8fb51d78d280/src/main/resources/diags.yml#L24-L30
which are read from the class path so I think we would have to either rebuild the support-diagnostics tool with different settings or inject a different configuration file into the JVM class path.
The other question is maybe: do we have hope that if we would wait longer the Kibana Diagnostics extraction would have been successful?
I'm not sure about the Kibana diagnostics part but we attempted an ES diagnostics (same API mode) and it was successful. The uncompressed size of the ES diagnostics is quite large (~660MB, with a 108MB cluster_state.json). I guess the failure could be related to the large size? But on the other hand, if the standalone diag-tool and the one in eck-diagnostics do the exact same thing, why was the different outcome?
Tried to capture an ECK diags - the command succeeded and we got the tar ball but it seems cluster diagnostic failed (the ECK dump part was correctly captured).
In
eck-diagnostic-errors.txt
in
eck-diagnostics.log
in kibana
diagnostics.log
Is there any other flag that we need to specify apart from
-o -r
?