[SOLR-5199] Restarting zookeeper makes the overseer stop processing queue events

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Duplicate
• Affects Version/s: 4.4
• Fix Version/s: 4.5.1, 4.6, 6.0
• Component/s: SolrCloud
• Labels: overseer zookeeper
• Assignee: Mark Miller
• Reporter: Jessica Cheng Mallet
• Created: 28/Aug/13 20:41
• Updated: 09/May/16 18:47
• Resolved: 11/Oct/13 16:38

Description

Taking the external zookeeper down (I'm just testing, so I only have one external zookeeper instance running) and then bringing it back up seems to have caused the overseer to stop processing queue event.

I tried to issue the delete collection command (curl 'http://localhost:7574/solr/admin/collections?action=DELETE&name=c1') and each time it just timed out. Looking at the zookeeper data, I see
...
/overseer
collection-queue-work
qn-0000000002
qn-0000000004
qn-0000000006
...
and the qn-xxx are not being processed.

Attached please find the log from the overseer (according to /overseer_elect/leader).

Issue Links

• duplicates: SOLR-5325 zk connection loss causes overseer leader loss
• relates to: SOLR-5325 zk connection loss causes overseer leader loss