janhoy opened a new pull request, #3956:
URL: https://github.com/apache/solr/pull/3956

   Another take on this. 
   
   While most runs pass, there is still flakiness and a new failure case we did 
not see before: `Caused by: java.io.EOFException: EOF reached while reading`... 
See 
[develocity](https://develocity.apache.org/scans/tests?search.startTimeMax=1765925999999&search.startTimeMin=1765753200000&search.timeZoneId=Europe%2FOslo&tests.container=org.apache.solr.client.solrj.impl.CloudHttp2SolrClientTest&tests.test=testHttpCspPerf).
   
   So I think this is because the common http clusterstate backend clients now 
do not refresh and thus have stale cluster data in the cache. But one crucial 
mistake I did right before committing the original fix was, after fixing some 
review comments, I added in an "improvement", to use `getRandomClient()` for 
collection creation and deletion, instead of the prior 
`cluster.getSolrClient()` call. While `cluster.getSolrClient()` is a zk-backed 
client that *do* refresh, the `getRandomClient()` clients are http CSP, and 
using these for collection creation and deletion will have polluted their 
(never updated) cache so it may send requests to a collection or node that does 
not exist. One of the test failures crash is for the deleteCollection call on 
line 346.
   
   There is also a `java.lang.AssertionError: expected:<2> but was:<3>` failure 
that I believe stems from having used the wrong client for collection business.
   
   So in this PR I switch back to `cluster.getSolrClient()` for collection 
creation and deletion. Will merge and see if it fixes all issues.
   
   I still have a fear that disabling CSP refresh for the random clients used 
by the other tests may also have side effects, but let's cross that bridge once 
it is proven.
   
   https://issues.apache.org/jira/browse/SOLR-18024
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to