Kibana can't load after upgrade to 5.6.0


(Tim Arp) #1

Hi,
After upgrading my cluster from 5.5.1 to 5.6.0, Kibana was the last piece to upgrade. On restart Kibana turned red and said "unknown setting" or something like that, I didn't record it at the time.
But the problem was this. The indices referenced in saved searches has changed from using the string "logstash-*" to an id like "AV6GbcQ9KD3D6TOqtyBc". Through some trial and error I figured this out and was able to manipulate a manual backup of my visualizations and searches by doing find and replace for the string representation of the index pattern. Was there supposed to be process to correct this for the move to 5.6? In order to complete the upgrade I needed to delete my .kibana index and restore my saved objects. But only after I went through them and reassigned the index pattern ids.

2nd part. Is the new ID used a unique generated ID that is only relevant for the cluster it's on? I ask this, because I would like to be able to export objects from my production cluster and load those onto my development cluster. Is the ID just a hash. This is really going to be a pain if it's unique to a cluster.

Thanks,
Tim


(Court Ewing) #2

Simply upgrading to 5.6 shouldn't have caused any issue because all existing saved objects (and their ids) remain untouched on upgrade. We did make a change to how ids were handled for new objects though, and there may be some rough edges that we need to smooth out.

Can you describe a bit more how you handled the upgrade? It doesn't sound like you did anything wrong, but we'd like to fix any of these issues, so specifics about how this was encountered are really helpful. Was Kibana red until you manually fixed the data? During your upgrade, did you recreate an index pattern or something?

The id is a unique generated ID by default, but you can still set an explicit id for the pattern via the advanced dropdown on the index pattern creation screen to support a dev/prod workflow like you're describing.


(Tim Arp) #3

Hi,
Since my post, I got an email that you guys released 5.6.1 which fixes this problem. So that is that. I wished I would have understood this better in my devl cluster before pushing 5.6.0 to prod. I just assumed that my devl cluster had some garbage in it that my prod one didn't have.
Thanks,
Tim


(system) #4

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.