We observed a situation recently where a large increase in the amount of 
data being ingested into ES caused log files in /var to fill to capacity on 
each node in a six node cluster.

Attempts to curl to a master VIP front-ended on the cluster would return a 
portion of the full json, but not all of it.  Imagine expecting ten "pages" 
of some type of information and only receiving 2-3 when it suddenly stops.

I first saw this when manually using curl on a Linux client.  I mistakenly 
thought it might be a client issue or a load balancer or proxy issue 
between me and the the cluster.

The next symptom came in the form of Elastic HQ being able to connect to 
the cluster but producing "*Error: Unable to Read Node List*".

The third symptom came from Kibana being unable to pull up any dashboards, 
timing out and proclaiming that our ES cluster was down.

Monitoring should have been enabled to monitor /var and we hadn't received 
any alerts on it, so of course we didn't think to stop and check any of the 
filesystems for out-of-space issues.  After we noticed this and corrected 
the problem, all our problems went away.

*The question is: Why would /var filling cause the cluster to return 
partial json in this manner?*

Thanks,

-- 
You received this message because you are subscribed to the Google Groups 
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/elasticsearch/45780f2b-df8d-4780-a107-bd8c6a58d325%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Reply via email to