mrmm / helm-charts Goto Github PK

@mrmm
I happened upon your kubernetes helm implementation for ES and was stumbling a little when trying to get it to work on 5-node Google Compute Platform. I was wondering if you might be able to point me in the right direction.

I started with a 3-node set-up, then extended it to 5-node one, running version 1.5.1 of kubernetes (both on master and on the nodes).

Given that PetSets have been replaced with StatefulSets, the only modification I made was to the elasticsearch/templates/elasticsearch-data-petset.yaml file.

from:

apiVersion: apps/v1alpha1
kind: PetSet

to:

apiVersion: apps/v1beta1
kind: StatefulSet

While the pv, pvc and services appear to get created properly, the Deployments and StatefulSets do not, with the errors like pod (es-002-client-1081588479-848hf) failed to fit in any node fit failure summary on nodes : MatchNodeSelector (5)

I tried scaling back replicas to 1 each in the elasticsearch/values.yaml .. and tried to tweak other memory limits downwards to try to get things working, but that didn't help, which suggests to me that perhaps other adjustments to get this working against version 1.5.1 might be needed.

Have you any thoughts on what might be going wrong here?

Also, if I was to try to get your implementation workings against v 5.x of ES, is the approach taken here, in terms of ensuring the vm.max_map_count=262144 gets set appropriately, heading in the right general direction?

For the dockerfile you use, is it just a case of adjusting the version number to 5.1.1 ... or do you envisage other adjustments might be necessary?

Thanks for your insights.