Once in a while (every few months), one of our clusters gets into a state where it freezes due to what appears to be endless indexing of design documents. When this occurs, view queries timeout and our REST server maxes out on connections to CBServer (since they're all hanging) which then kills the REST server. Also, cbbackup seems to fail when the server gets into this state.
Currently it's not clear what causes it. We've resolved it in the past by flushing the bucket restoring a backup from the previous day, but that's not an ideal approach. This solution would also suggest that somehow a newly added document would be the cause, but if there was such an erroneous document then our mobile apps (which use CBLite) would start acting strange as well once they sync that document, no? A few questions then: 1. What might cause endless indexing? cbcollect_info reveals some map errors but no reduce errors. 2. cbbackup -vvv doesn't reveal any info about why backup isn't finishing during the index choke. Is there another way to get more info? 3. Is there another way to perform a data backup from a cluster in case of this kind of cluster-wide freeze? -- You received this message because you are subscribed to the Google Groups "Couchbase" group. To unsubscribe from this group and stop receiving emails from it, send an email to couchbase+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
