[SOLR-12544] ZkStateReader can cache deleted collections and never refresh it

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 7.2.1
• Assignee: Unassigned
• Reporter: Varun Thacker
• Created: 11/Jul/18 16:30
• Updated: 21/Nov/19 00:41
• Resolved: 08/Jan/19 19:14

Description

After a delete collection call, CLUSTERSTATUS starts breaking with this error permanently with this error 

org.apache.solr.common.SolrException: Error loading config name for collection my_collection
    at org.apache.solr.common.cloud.ZkStateReader.readConfigName(ZkStateReader.java:198)
    at org.apache.solr.handler.admin.ClusterStatus.getClusterStatus(ClusterStatus.java:141)
...
Caused by: org.apache.zookeeper.KeeperException$NoNodeException: KeeperErrorCode = NoNode for /collections/my_collection
...

SOLR-10720 addresses the problem by skipping over the collection as it was aimed to fix the  race condition between delete collection and CLUSTERSTATUS being called.

 

The fact that we see the error never go away , means there is another bug lingering which will make the state never refresh and thus other calls list LIST will always show the collection. 

 

This happened with Solr 7.2.1 and doesn't happen very often. But when it does the only solution is to restart the node. 

Issue Links

• is related to: SOLR-9440 ZkStateReader on a client can cache collection state and never refresh it
• is related to: SOLR-10720 Aggressive removal of a collection breaks cluster state