dsmiley commented on PR #5358:
URL: https://github.com/apache/solr/pull/5358#issuecomment-7057306154

   🤖 Heads-up: this change appears to have broken the nightly 
[Solr-TestIntegration-main](https://ci-builds.apache.org/job/Solr/job/Solr-TestIntegration-main/)
 job (BATS). Builds 
[733](https://ci-builds.apache.org/job/Solr/job/Solr-TestIntegration-main/733/) 
and 
[734](https://ci-builds.apache.org/job/Solr/job/Solr-TestIntegration-main/734/) 
fail the same 3 tests on two different agents; 
[731](https://ci-builds.apache.org/job/Solr/job/Solr-TestIntegration-main/731/) 
(at 4890215acc8, just before this commit) passed. Build 732 never ran its tests 
because of an agent JAVA_HOME problem. Of the commits in that range, this is 
the only one that touches `bin/solr` start behavior. Until it's fixed, this may 
also hold up backports of this change and others, since the integration suite 
is red.
   
   **Failing tests**
   - `test_secure_example.bats:39`: secure example starts two HTTPS nodes with 
authentication and ZooKeeper ACLs
   - `test_ssl.bats:122`: use different hostname when not checking peer-name 
(`solr restart` failed)
   - `test_ssl.bats:360`: start solr with mTLS needed (`solr start -z 
localhost:${ZK_PORT} -p ${SOLR3_PORT}` failed)
   
   **Analysis**
   
   `start_solr` used to wait only until `lsof` saw the port listening. It now 
runs `status -p $SOLR_PORT_LISTEN --max-wait-secs $SOLR_START_WAIT`, which 
needs a successful request to `https://localhost:<port>/solr`. `waitForSolrUp` 
only treats 401/403 as "up". Any other error is retried until the timeout, and 
then `start` returns 1.
   
   - **test_ssl.bats:122 and :360**: both tests deliberately start a node that 
the CLI client can't talk to.
     - :122 sets `-Dsolr.jetty.ssl.sni.host.check.enabled=true`, so a request 
to `localhost` gets "Invalid SNI" (HTTP 400).
     - :360 sets `SOLR_SSL_CLIENT_HOSTNAME_VERIFICATION=true`. The test's own 
comment says "We can't check if the server has come up, because we can't 
connect to it".
     - In both cases the node does come up (the log shows cores loading), but 
`start` waits the full 180s and then fails.
   - **test_secure_example.bats:39** (less certain): step 5 restarts node1 with 
ZK digest ACLs, and that restart failed after about 2s instead of waiting out 
the 15s timeout. The `solr.log` tail that was printed comes from the previous 
process's shutdown. That suggests `status -p` exited right away rather than 
polling. The likely path is `processForPort()` returning empty ("Could not find 
a running Solr on port …"), maybe because the new JVM wasn't visible yet. The 
`runTool` port branch exits 1 at that point without honoring `--max-wait-secs`. 
The real message is hidden because `wait_for_solr_ready` prints `status_output` 
only with `--verbose`.
   
   **Possible fixes**
   1. In `waitForSolrUp`, treat any HTTP response from the server (e.g. Jetty's 
400 for the SNI case) as "up", as 401/403 already are. A TLS rejection could 
arguably count too. The alternative is to change the two SSL tests to tolerate 
a failing `start` (`run solr start` …).
   2. With `--max-wait-secs`, make `status -p` keep polling for the process 
instead of exiting right away when none is found yet.
   3. Print `status_output` on failure regardless of `--verbose`, so CI logs 
show the cause.
   4. The CATS ports merged in #5323 (`SslTest`, `SecureExampleTest`) probably 
have the same expectations.
   
   **Testing:** run the BATS tests locally with `./gradlew iTest --tests 
test_ssl.bats` and `./gradlew iTest --tests test_secure_example.bats`, or run 
the CATS equivalents with `./gradlew -p solr/packaging catsTests --tests 
"*SslTest"`.
   
   🤖 Generated with [Claude Code](https://claude.com/claude-code)
   


-- 
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