Add option to mount existing Amazon EFS volume. #17
Conversation
| # | ||
| # You can create a new EFS volume using the CLI: | ||
| # $ aws efs create-file-system --creation-token $(uuidgen) | ||
| #elasticFileSystem: fs-47a2c22e |
There was a problem hiding this comment.
like hostedZoneId, routeTableId, etc., should we name this elasticFileSystemId? (It is specified via IDs according to e.g. http://docs.aws.amazon.com/cli/latest/reference/efs/create-file-system.html
There was a problem hiding this comment.
Agreed. Will fix this.
|
I'm investigating some issues regarding service restart and am considering changing this to a full-fledged systemd mount (but that has side-effects too because i need to template the NFS path for the availability zones). Please hold off on merging this. Thanks |
|
Sure! |
Current coverage is 56.90% (diff: 100%)
|
|
Not so difficult, it was just past my bedtime and i wanted to make sure i got it right this time. 🤓 Verified to be working now! If there is sufficient demand and i have enough time i'll write some code to create the EFS volume as well, but i think the |
|
I agree that we should add the EFS volume creation part when we have sufficient demand. |
…w-kube-aws to hcom-flavour * commit '13aa1d1b0569f21304545c005c5da9f810a41b66': (41 commits) Continue to allow `system:nodes` to act as `system:node-proxier` Rename cluster, user, context to 'default' Use localhost instead of 127.0.0.1 Add comment Use service account token in master kube-proxies Address review points Convert kube-proxy to a DaemonSet Add in missing ASG IAM permission for experimental.nodeDrainer. Support deployment to AWS GovCloud Update kubedns Only add CIDR from NLB-backed endpoints to controller SG Fixes kubernetes-retired#946 Support drop-ins in customSystemdUnits Remove tests for constraints that are no longer true Add support for apiAccessAllowedSourceCIDRs to NLB-backed endpoints Use correct predicate function Fix logic to check whether the load balancer is a NLB Add support for NLB-backed API endpoints Bump default k8s to 1.7.6 Fixes kubernetes-retired#942 etcdadm cannot find ca.pem Remove `createRecordSet` to help newcomers bring up their first clusters (kubernetes-retired#936) ...
This still needs some work but i think it currently qualifies and fills a need.
Creating (and later finding) an EFS volume through cloudformation is more trouble than it's worth so the option currently requires an existing EFS-id. We can later turn this into an option which requires a unique id for creating the volume. We'd need to create the volume before we evaluate the stack-template though and my Go skills are still quite rusty so i think it's better to do that in a separate PR.
(nearly) fixes #10