Elasticsearch unavailable during upgrade


(Josh Cummings) #1

We are in the process of upgrading from 1.6 to 2.1. The logs show a lot of activity with regard to upgrading each index; however, during that time, connections on 9200 are refused. Is that expected behavior?


(Mark Walkom) #2

What sort of connections? What state is the cluster in?


(Josh Cummings) #3

http connections. https://localhost:9200 returns a connection refused.

Typically, there is a message in the logs like "[servername] publish_address {ipaddress:9200}, bound_address{[::]:9200}"

I don't see this in the logs during the upgrade, so I suppose this is why 9200 isn't responding. It just seems incorrect to not see that message since there is an upgrade-status url that one can ostensibly hit while the system is upgrading.

When you ask "what state", I can't query the red/yellow/green state since connections are refused on 9200, or is there another way for me to query that?


(Mark Walkom) #4

Take a look at Connectivity issues with a new/upgraded 2.X cluster? Read here first :)


(Josh Cummings) #5

I'm familiar with the document. When there are no indices to upgrade, I am able to hit the cluster just fine because I've already set network.host to 0.0.0.0. It is during the index upgrade that I can't get a response. Thus far, I've tried 0.0.0.0 and non_loopback.

I will try some other values for network.host to see what kind of behavior I get. If my ES 2.1 cluster is pointed at a data mount with no indices to upgrade, 9200 responds fine.


(system) #6