[COUCHDB-1224] Couch crash after View compaction, unresponsive for 1-2 mins, view reindexes from scratch

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Won't Fix
• Affects Version/s: 1.0.2
• Component/s: Database Core
• Labels: compaction corruption view

Environment

CentOS 5.6 Server, on XFS raid1 array. Running as user couchdb. Build was 1.0.3aaf28707-git - a patched version of 1.0.2 for COUCHDB-885

• Assignee: Unassigned
• Reporter: Alex Markham
• Created: 17/Jul/11 13:20
• Updated: 02/Sep/18 06:33
• Resolved: 02/Sep/18 06:33

Description

A largish couchdb database was running through its view compactions (we run 3 times a week) when it stopped responding to calls. I only realised after the event when the logs of queries to couch showed very long response times.

The last thing in the logs is: "View index compaction complete for <View Name>" and then a bunch of stack traces, lasting a couple of minutes.

Couch is now reindexing the whole of that <View Name> which will take hours
Perhaps because it got corrupted during the compaction?

eg.
[Sun, 17 Jul 2011 12:07:09 GMT] [info] [<0.30126.2013>] checkpointing view update at seq 1 for <DB> <View Name>
[Sun, 17 Jul 2011 12:07:09 GMT] [info] [<0.13656.133>] checkpointing view update at seq 2103372 for <DB> <Another view>

Issue Links

• is related to: COUCHDB-994 Crash after compacting large views