[SOLR-8335] HdfsLockFactory does not allow core to come up after a node was killed

• Type: Bug
• Status: Reopened
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 5.0, 5.1, 5.2, 5.2.1, 5.3, 5.3.1
• Component/s: Hadoop Integration, hdfs
• Assignee: Mark Miller
• Reporter: Varun Thacker
• Created: 24/Nov/15 04:26
• Updated: 21/Nov/19 00:41

Description

When using HdfsLockFactory if a node gets killed instead of a graceful shutdown the write.lock file remains in HDFS . The next time you start the node the core doesn't load up because of LockObtainFailedException .

I was able to reproduce this in all 5.x versions of Solr . The problem wasn't there when I tested it in 4.10.4

Steps to reproduce this on 5.x

1. Create directory in HDFS : bin/hdfs dfs -mkdir /solr
2. Start Solr: bin/solr start -Dsolr.directoryFactory=HdfsDirectoryFactory -Dsolr.lock.type=hdfs -Dsolr.data.dir=hdfs://localhost:9000/solr -Dsolr.updatelog=hdfs://localhost:9000/solr
3. Create core: ./bin/solr create -c test -n data_driven
4. Kill solr
5. The lock file is there in HDFS and is called write.lock
6. Start Solr again and you get a stack trace like this:

2015-11-23 13:28:04.287 ERROR (coreLoadExecutor-6-thread-1) [   x:test] o.a.s.c.CoreContainer Error creating core [test]: Index locked for write for core 'test'. Solr now longer supports forceful unlocking via 'unlockOnStartup'. Please verify locks manually!
org.apache.solr.common.SolrException: Index locked for write for core 'test'. Solr now longer supports forceful unlocking via 'unlockOnStartup'. Please verify locks manually!
        at org.apache.solr.core.SolrCore.<init>(SolrCore.java:820)
        at org.apache.solr.core.SolrCore.<init>(SolrCore.java:659)
        at org.apache.solr.core.CoreContainer.create(CoreContainer.java:723)
        at org.apache.solr.core.CoreContainer$1.call(CoreContainer.java:443)
        at org.apache.solr.core.CoreContainer$1.call(CoreContainer.java:434)
        at java.util.concurrent.FutureTask.run(FutureTask.java:266)
        at org.apache.solr.common.util.ExecutorUtil$MDCAwareThreadPoolExecutor$1.run(ExecutorUtil.java:210)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
        at java.lang.Thread.run(Thread.java:745)
Caused by: org.apache.lucene.store.LockObtainFailedException: Index locked for write for core 'test'. Solr now longer supports forceful unlocking via 'unlockOnStartup'. Please verify locks manually!
        at org.apache.solr.core.SolrCore.initIndex(SolrCore.java:528)
        at org.apache.solr.core.SolrCore.<init>(SolrCore.java:761)
        ... 9 more
2015-11-23 13:28:04.289 ERROR (coreContainerWorkExecutor-2-thread-1) [   ] o.a.s.c.CoreContainer Error waiting for SolrCore to be created
java.util.concurrent.ExecutionException: org.apache.solr.common.SolrException: Unable to create core [test]
        at java.util.concurrent.FutureTask.report(FutureTask.java:122)
        at java.util.concurrent.FutureTask.get(FutureTask.java:192)
        at org.apache.solr.core.CoreContainer$2.run(CoreContainer.java:472)
        at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
        at java.util.concurrent.FutureTask.run(FutureTask.java:266)
        at org.apache.solr.common.util.ExecutorUtil$MDCAwareThreadPoolExecutor$1.run(ExecutorUtil.java:210)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
        at java.lang.Thread.run(Thread.java:745)
Caused by: org.apache.solr.common.SolrException: Unable to create core [test]
        at org.apache.solr.core.CoreContainer.create(CoreContainer.java:737)
        at org.apache.solr.core.CoreContainer$1.call(CoreContainer.java:443)
        at org.apache.solr.core.CoreContainer$1.call(CoreContainer.java:434)
        ... 5 more
Caused by: org.apache.solr.common.SolrException: Index locked for write for core 'test'. Solr now longer supports forceful unlocking via 'unlockOnStartup'. Please verify locks manually!
        at org.apache.solr.core.SolrCore.<init>(SolrCore.java:820)
        at org.apache.solr.core.SolrCore.<init>(SolrCore.java:659)
        at org.apache.solr.core.CoreContainer.create(CoreContainer.java:723)
        ... 7 more
Caused by: org.apache.lucene.store.LockObtainFailedException: Index locked for write for core 'test'. Solr now longer supports forceful unlocking via 'unlockOnStartup'. Please verify locks manually!
        at org.apache.solr.core.SolrCore.initIndex(SolrCore.java:528)
        at org.apache.solr.core.SolrCore.<init>(SolrCore.java:761)
        ... 9 more

In 4.10.4 I saw these two differences

1. The lock file name was different . It's something like : /solr/index/HdfsDirectory@46ad6bd3 lockFactory=[e-mail]
2. When the node is started again after it was killed , it loaded up the core just fine but there were two lock files in hdfs now . 4b44b5f6-write.lock is the latest one

/solr/index/HdfsDirectory@46ad6bd3 lockFactory=[e-mail]
/solr/index/HdfsDirectory@52959724 lockFactory=[e-mail]

Issue Links

• is duplicated by: SOLR-9024 When shutdown solr process it don't delete write.lock on HDFS
• relates to: SOLR-8169 Need LockFactory impl that uses ZooKeeper as replacement for HdfsLockFactory
• links to: GitHub Pull Request #471
• links to: GitHub Pull Request #471