[SOLR-4608] Update Log replay should use the default processor chain

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 4.1, 4.2
• Fix Version/s: 4.2.1, 4.3, 6.0
• Component/s: SolrCloud
• Assignee: Yonik Seeley
• Reporter: Ludovic Boutros
• Created: 18/Mar/13 16:00
• Updated: 09/May/16 18:48
• Resolved: 21/Mar/13 18:08

Description

If a processor chain is used with custom processors,
they are not used in case of node failure during log replay.

Here is the code:

UpdateLog.java

    public void doReplay(TransactionLog translog) {
      try {
        loglog.warn("Starting log replay " + translog + " active="+activeLog + " starting pos=" + recoveryInfo.positionOfStart);

        tlogReader = translog.getReader(recoveryInfo.positionOfStart);

        // NOTE: we don't currently handle a core reload during recovery.  This would cause the core
        // to change underneath us.

        // TODO: use the standard request factory?  We won't get any custom configuration instantiating this way.
        RunUpdateProcessorFactory runFac = new RunUpdateProcessorFactory();
        DistributedUpdateProcessorFactory magicFac = new DistributedUpdateProcessorFactory();
        runFac.init(new NamedList());
        magicFac.init(new NamedList());

        UpdateRequestProcessor proc = magicFac.getInstance(req, rsp, runFac.getInstance(req, rsp, null));

I think this is a big issue, because a lot of people will discover it when a node will crash in the best case... and I think it's too late.

It means to me that processor chains are not usable with Solr Cloud currently.