Destroying deployments fail

[erikgraa]

Hello

I have a problem with my vRA 7.5.0 (build 10053500) instance where I'm unable to destroy deployments

Deployments that fail during deploying also fail to clean up, leaving machines

Deployments that have failed destroying themselves only have the "dismiss" and/or "resubmit" actions, so I can't forcibly destroy them afterwards

The "destroy" task on VMs in have the following error:

"Internal error in processing component request: [Rest Error]: {Status code: 400}. {Error code: 10101}, {Error Source: null}, {Error Msg: Invalid argument.}, {System Msg: No matching constant for [0]}

This is leaving me with a ton of machines that have no idea how to remove and I'm soon out of compute so I can't test new deployments.

Any tips?

[GayathriS]

Hi

When VM deployment fails it basically deletes the VM , if VM's are left over then check if below settings are setted to false.

1. Go to the installation directory (C:\Program Files (x86)\VMware\vCAC\Agents\);
2. Go to the folder where the name of the agent is equal to the name of the folder that contains the agent;

Check the current configuration settings by running the following command.

‘ .\DynamicOps.Vrm.VRMencrypt.exe VRMAgent.exe.config get‘

From the output  " doDeletes " property should  be setted to TRUE , if thats the case then it is expected to Delete VM's post deployment failure.

-------------------

For the destroy task which is not completeing : try to do a force destroy and check if that succedds

Force Destroy a Deployment After a Failed Destroy Request

Please mark this as "correct" or " Helpfull" if this answers your query.

Regards

Gayathri

[erikgraa]

Thanks for answering,

1) I haven't got debug mode on, so VMs definitely /should/ be destroyed on a failed deployment

2) The "force destroy" option is gone after a deployment has failed, and the "action wheel" only contains dismiss and resubmit.

How can I forcibly destroy a deployment from CLI if it's not possible through GUI?

How can I remove the VMs? Unregister them in vRA and then remove from vCenter? Will the compute reservation reflect the new available compute/storage?

[daphnissov]

If VMs get built but don't show in a deployment, go to Infrastructure => Managed Machines and delete them there. For forcibly killing deployments, you'll need to use CloudClient CLI tool which you can get from the vRA Tools page.

[erikgraa]

They get built and they do show in deployments, but don't go away if a deployment fails or if I try to destroy whatever constituents a deployment has.

I'm stuck with 20 deployments, about 100 VMs and associated NSX-created port groups/edges by now.

[daphnissov]

First thing I'd make sure is that you have the latest hot fix for 7.5 applied.

[erikgraa]

I'll roll that out tomorrow and try, you mean this one right?

VMware Knowledge Base

I have already unregistereed a number of VMs that are part of failed deployments -- can I just delete them from vcenter?

[daphnissov]

Yes, that's the KB. You can delete those VMs from vCenter if you've unregistered them, yes. Depending on what extensibility you had fire during the provisioning process, though, you will need to roll that back as well.

[erikgraa]

I've updated to the patch in VMware Knowledge Base  now, and it was successful

However, checking out existing deployments now, failed and successful, the VMs and components don't even load, and if I press "Destroy", I get "the form initilization failed" and I can't press "submit".

Even pressing publish on a blueprint will cause the UI to load for minutes.

[daphnissov]

I'd say it's time to open that SR.

[erikgraa]

Seems I understand why the blueprint wouldn't publish -- the Cumulative Update has destroyed all of my network profiles it seems.

[erikgraa · Accepted solution]

Re. the network profiles that weren't available: the SQL server associated with the IAAS server had failed to start after a boot.

I've been able to forcibly unregister all machines using the vRA Cloud Client. After doing so I deleted them from vCenter.

The only remainder seems to be about 50 virtual port groups spun up by NSX that have failed to be destroyed. I'll just leave them be as this is just a test environment.

Thanks!