[jira] [Commented] (SOLR-13028) Harden AutoAddReplicasPlanActionTest#testSimple

["Hoss Man (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SOLR-13028?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16710392#comment-16710392 ] 

Hoss Man commented on SOLR-13028:
---------------------------------

So where exactly is the bug? Are there multiple bugs?
 * All the code in question running on the solr nodes is using the CloudSolrClient that comes from the SolrClientCloudManager (from the ZkController) ... since every solr node has ZK watches for the live nodes (and state.json for each collection) shouldn't each ZkController be able to purge it's CloudSolrClient's connection pool if we know a node has gone down ?

 * Sprinkled throughout the solr code base, we have a lot of places where {{NoHttpResponseException}} is checked for explicitly and considered a "communication error" that can be retried ... is that appropriate here in the AutoScalling Policy/Snitch logic?
 ** Other methods in SolrClientNodeStateProvider have explicit checks for {{cause instanceof NoHttpResponseException}} when encountering an exception in which case they retry, but SolrClientNodeStateProvider's static inner class AutoScalingSnitch doesn't – it has nearly cut/pasted code that only looks for SocketException ... should this code path be retrying on NoHttpResponseException?
 ** Ironically the error it throws if it retries on SocketException multiple times w/o succeeding missleadingly claims it encountered a NoHttpResponseException – suggesting blatent cut/paste abuse...
{code:java}
        while (cnt++ < retries) {
          try {
            rsp = snitchContext.invoke(solrNode, CommonParams.METRICS_PATH, params);
          } catch (SolrException | SolrServerException | SocketException e) {
            boolean hasCauseSocketException = false;
            Throwable cause = e;
            while (cause != null) {
              if (cause instanceof SocketException) {
                hasCauseSocketException = true;
                break;
              }
              cause = cause.getCause();
            }
            if (hasCauseSocketException || e instanceof SocketException) {
              log.info("Error on getting remote info, trying again: " + e.getMessage());
              Thread.sleep(500);
              continue;
            } else {
              throw e;
            }
          }
        }
        
        if (cnt == retries) {
          throw new SolrException(ErrorCode.SERVER_ERROR, "Could not get remote info after many retries on NoHttpResponseException");
        }
{code}

> Harden AutoAddReplicasPlanActionTest#testSimple
> -----------------------------------------------
>
>                 Key: SOLR-13028
>                 URL: https://issues.apache.org/jira/browse/SOLR-13028
>             Project: Solr
>          Issue Type: Sub-task
>      Security Level: Public(Default Security Level. Issues are Public) 
>            Reporter: Mark Miller
>            Priority: Major
>         Attachments: sarowe__Lucene-Solr-BadApple-tests-master__229.log.txt
>
>




--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org