Validation failure on gp2 volume type for node pools #1170
Comments
@shraykay Thanks for the concise report! Your fix does seem to meet the body! |
We ran into the same issue ourselves. I was about to reproduce independently. |
@shraykay I was going to submit a PR, but if you're already a contributor you can probably get it done faster. Ping me if you plan on doing this in the near future, otherwise I'll give it a shot. Can @mumoshu or @shraykay point me where a test for this ought to go? (the code is simple but I'm out of practice with Go and don't know what the current testing layout best practices are) |
results in output:
|
don't set it to 0, just comment out the entire line for the iops |
Note that commenting out the entire line is not preferable, as it throws away the useful validation when the volume type is
|
That makes sense, I'm just validating something with the AWS api before dropping a PR. Thanks! |
@luck02 Thanks for your support! Tests should go into |
Just as an amusing point of reference: So the test needs tweaking, ahh well. that was confusing for awhile. |
PR here: #1171 tests pass locally but I didn't run e2e, I'm assuming the build infra does that? |
…retired#1171) Improved tests and code to respect changes in EC2 validate volume API. We can no longer submit IOPS 0 as being synomous with unset field / null. In `core/nodepool/cluster.go`, the function: `validateWorkerRootVolume` is changed to call an another function to generate EC2 volume parameters. Also tweaked tests around that. Fixes kubernetes-retired#1170
Is there an ETA for this fix to be released generally? I am not able to deploy my cluster using |
@zachrattner i believe this was merged into master in #1171 |
Thanks @shraykay I see the latest release was back in December (https://github.com/kubernetes-incubator/kube-aws/releases) and this fix was within the last month. The readme mentions that master might not be stable and we should pull from release branches if we need a stable build. Where should I pull from to get a stable codebase with the fix? |
master is stable IMO, and works with kubernetes 1.9 as well. if you are especially worried you can cherry pick the commit into a release and compile it. |
Is this broken again? last I checked AWS rolled back this change. We may not have noticed as we started using a locally baked kube-aws binary that had this baked in. Let us know as we'll ping our AWS Technical Account Managers @zachrattner, also good to know for general knowledge. |
@luck02 Yes, I'm seeing this issue on v0.9.9. Keep me posted! I'll try cherry picking the fix on top of the 0.9.9 release in the meantime |
For anyone else running into this issue with the same problem of not having a stable build with the fix present, I can confirm the deployment with
Helpful background info: https://kubernetes-incubator.github.io/kube-aws/guides/developer-guide.html |
…retired#1171) Improved tests and code to respect changes in EC2 validate volume API. We can no longer submit IOPS 0 as being synomous with unset field / null. In `core/nodepool/cluster.go`, the function: `validateWorkerRootVolume` is changed to call an another function to generate EC2 volume parameters. Also tweaked tests around that. Fixes kubernetes-retired#1170
…retired#1171) Improved tests and code to respect changes in EC2 validate volume API. We can no longer submit IOPS 0 as being synomous with unset field / null. In `core/nodepool/cluster.go`, the function: `validateWorkerRootVolume` is changed to call an another function to generate EC2 volume parameters. Also tweaked tests around that. Fixes kubernetes-retired#1170
…retired#1171) Improved tests and code to respect changes in EC2 validate volume API. We can no longer submit IOPS 0 as being synomous with unset field / null. In `core/nodepool/cluster.go`, the function: `validateWorkerRootVolume` is changed to call an another function to generate EC2 volume parameters. Also tweaked tests around that. Fixes kubernetes-retired#1170
it seems that AWS has changed the way their cloudformation validation occurs. when validate is ran, this error occurs during the dry-run if you're using disk type = gp2. after commenting this out in and compiling a binary myself I was able to pass validation: cluster.go line 203
The text was updated successfully, but these errors were encountered: