Ubuntu 14.04 upgrade process for 1.4 to 1.5 and mini-ish FAQ


#1

My upgrade process for Ubuntu 14.04 from a working 1.4 install:

Comment out:
deb http://packages.elasticsearch.org/elasticsearch/1.4/debian stable main
deb http://packages.elasticsearch.org/logstash/1.4/debian stable main

and add in:
deb http://packages.elastic.co/elasticsearch/1.5/debian stable main
deb http://packages.elastic.co/logstash/1.5/debian stable main

sudo service logstash stop
sudo service elasticsearch stop

sudo apt-get update
sudo apt-get dist-upgrade

sudo service elasticsearch start
sudo service logstash start

Download kibana for you architecture at:
https://www.elastic.co/downloads/kibana

Mini-ish FAQ
Kibana is it's own beast now instead of integrating with a web server, I extracted it /opt/kibana
Your current Kibana 3 dashboard and queries don't port over, so be prepared to do a lot of rework
I had to increase my ES_HEAP_SIZE to 1 gig when doing the initial Kibana 4 discovery otherwise it would fail

That's it for now. I'm sticking with Kibana 3 for a while. Currently I'm not sure this upgrade was even worth it.


(Mark Walkom) #2

Thanks for the write up :slight_smile:


(Jason Wee) #3

this documentation definitely is really helpful. at times, doing automatic upgrade is always desirable and especially in this face pace life. just a question, do you feel safe or if the automatic upgrade is done by third party? i mean as important as a datastore if not upgraded successfully, in production, it would mean a disaster. maybe that's just my little knowledge, if you have better views, could you write more by elaborate on how to successfully upgrade elasticsearch?


#4

I do not do automatic updates. My production ELK stack is virtualized, so when it comes time for updating I make a snapshot of the VM, and then run the updates and test. I also have a development box to test on, this too gets updated before I even think about the production one. If I ever planned to do automatic updates, I would at least schedule them while I'm at work on site...so if something tanks I would be there and active ready to respond, not have something broken from midnight until I got in for example.


#5

Last word on this. After restarting the elasticsearch service a few times I started to get:

marking and sending shard failed due to [failed recovery]

errors and data stopped going into ES. Long story short I blew out the data and downgraded back to ES/Logstash 1.4.5. 1.5 just is not ready for primtime in my experience.


(system) #6