Lost index after Cluster fiddling


(Marc Fielding) #1

Hi Everyone,

I've recently started out with Elastic and had everything running absolutely perfectly on my local dev machine.

Yesterday I thought I'd get an idea of how clusters work, so I installed Elastic on my Raspberry Pi 2, gave it an IP address, the same cluster name and a proper node name.

I noticed connect messages on the console of my windows machine, in Marvel although the little summary at the top showed 2 nodes in the node list it only ever showed the windows machine and both nodes. I used the _cluster endpoint to investigate and it appeared they'd connected ok.

This morning I wake up, sit down and clicked refresh on a PHP API I was working on(using elastic) and got "No nodes available" despite the one on my local machine apparently being up and green AND the other on the Pi!

After a couple of restarts on both the Pi and the windows box I was still getting the same error so I decided to stop the node on the Pi and switch the configuration of the windows machine back to it's localhost settings(network host to 127.0.0.1).

Now the index I created has completely disappeared!

  1. What did I do wrong?

  2. Is there any way to fix it?

Kind Regards

Marc


(Marc Fielding) #2

The first issue I couldn't figure out so I overwrote my elastic directory with a fresh download, set the cluster and node name. Fired up under windows and then on my Pi, they're now talking just fine.

The problem with data - Obviously when you change the cluster name Elastic sets up a new directory for the data, so I had elastic-marc.dev yesterday and when things started breaking I changed it to elastic.

I still don't know what caused the problem overnight when everything seemed to get corrupted but will see what happens today!

Regards

Marc


(Mark Walkom) #3

There should be evidence of something in the logs, GC causing node timeouts or something else.


(system) #4