Home | Forums | Couchbase | Couchbase Server 2.0

Couchbase-elasticsearch sync issue

2 replies [Last post]
Fri, 10/05/2012 - 06:22
amycouchbase
Offline
Joined: 10/05/2012
Groups: None

I've been investigating whether to use couchbase with elasticsearch. There's an excellent blog http://blog.couchbase.com/couchbase-and-full-text-search-couchbase-trans... on how to set this up. I have one issue.
If elasticsearch is down, and you send a document to couchbase, then bring back up elasticsearch, the document won't be indexed by elasticsearch.
I've tried flushing the index to have elasticsearch pick up the document, but flushing doesn't seem to work. In fact, I see that only half the shards flushed successfully.
curl -XPOST http://localhost:9200/beer-sample/_flush
{"ok":true,"_shards":{"total":10,"successful":5,"failed":0}}
I'm not sure if this is an issue as half my shards seem to be under the name "Unassigned".
My only current solution is to restart Couchbase, but obviously I wouldn't want to do that in production.
Am I missing something?

Top
Mon, 10/08/2012 - 15:10
mschoch
Offline
Joined: 08/03/2011
Groups: None

Hi Amy,

Just saw this post here and wasn't sure if it was before or after our email exchange about this. The XDCR process used to send data to the Couchbase ElasticSearch adapter follows an exponential backoff process when retrying connections to an ElasticSearch cluster that was down. Depending on how long your ES cluster was down, this could cause significant delays before those changes are propagated.

I'm still working to reproduce the scenario you described and get clarification on the parameters of the backoff. I'll update you when I know more.

marty

Top
Fri, 10/12/2012 - 06:21
amycouchbase
Offline
Joined: 10/05/2012
Groups: None

I realise that what I had done was delete the index, and re-add it again. To get my documents re-indexed I needed to restart couchbase. I'm informed that if you want to delete an index you should first delete the XDCR transfer.

Top