You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@cloudstack.apache.org by Venkata Siva Vijayendra Bhamidipati <vi...@citrix.com> on 2013/08/01 01:41:34 UTC

Re: Review Request 13093: Put retry logic to check existence/non-existence of created/deleted files even though vSphere returns successful completion on wait for asynchronous task.


> On July 31, 2013, 1:49 a.m., Jenkins Cloudstack.org wrote:
> > Review 13093 PASSED the build test
> > The url of build cloudstack-master-with-patch #90 is : http://jenkins.cloudstack.org/job/cloudstack-master-with-patch/90/

When speaking with Kelven regarding a different bug, Kelven pointed out that waitForTask() is a cloudstack wrapper and its implementation may be causing the premature return of success for the API - this is more plausible since many places in the code are now running into this issue. Will look into this and reupload the patch.


- Venkata Siva Vijayendra


-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/13093/#review24312
-----------------------------------------------------------


On July 31, 2013, 12:41 a.m., Venkata Siva Vijayendra Bhamidipati wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/13093/
> -----------------------------------------------------------
> 
> (Updated July 31, 2013, 12:41 a.m.)
> 
> 
> Review request for cloudstack, Alena Prokharchyk, Chip Childers, Kelven Yang, and Sateesh Chodapuneedi.
> 
> 
> Bugs: CLOUDSTACK-3568
> 
> 
> Repository: cloudstack-git
> 
> 
> Description
> -------
> 
> When a large number of VMs are created in parallel in vmware deployments, vSphere can slow down and return wrong results for its APIs. For example, the mgmt server correctly waits on the results of asynchronous tasks like deletion/moving of VM root volume files, but vSphere is returning success for those APIs even when the underlying files aren't yet removed/created from/on the actual storage. So we need to work around this using retry logic. This fix puts that in when creating VMs.
> 
> 
> Diffs
> -----
> 
>   plugins/hypervisors/vmware/src/com/cloud/storage/resource/VmwareStorageProcessor.java 112a0cb 
> 
> Diff: https://reviews.apache.org/r/13093/diff/
> 
> 
> Testing
> -------
> 
> The fix has been tested with parallel creation of ~15 VMs on vmware and the described issues have not been seen. There seem to be some other exceptions with some VMs, but they are not related to this code path.
> 
> 
> Thanks,
> 
> Venkata Siva Vijayendra Bhamidipati
> 
>