Attempt to deploy fresh 4.20.1 utilizing xcp-ng 8.3 fails to deploy system VMs

[alsko-icom]

cloudstack4.20.1.alsko.log

Hello team,
We decided to test out latest xcp-ng 8.3 hosts that are currently also visible through an XO tool. We waited for the Cloudstack version to support it and tried the required steps today. Our previous test with 4.20 was using KVM and we did not like the cluster HA delays we witnessed. XCP-ng seems a bit more promising. It seems that the process to deploy the system VMs failed because the XCP-ng hosts cannot mount the nfs share required. Attaching a log for your reference, let me know if you can help with what we are doing wrong.

[DaanHoogland]

@alsko-icom , the logs are not visible.

[alsko-icom]

cloudstack4.20.1.alsko.txt reuploaded file with txt extension.

[DaanHoogland]

it is now zero bytes, @alsko-icom . can you create a gist or use some other upload/ftp service?

[alsko-icom]

added the file to my personal nextcloud server and as a gist too.
https://gist.github.com/alsko-icom/c835b6f4ad87a27bd3ca96604dcf9c4c

and the nextcloud
https://secure.you-dream.com/index.php/s/PeYpjtBw9fJKHfm

[Pearl1594]

@alsko-icom can you please confirm if the systemvm template has been registered / seeded?

[DaanHoogland]

@alsko-icom please checkt the template as @Pearl1594 suggests.

Why do you think the mount of the nfs share is the problem?

[alsko-icom]

pasting here screenshot of the seed command for your reference

The team noticed some errors during deployment regarding NFS SR being unable to mount on the xcp-ng hosts. So if nfs mount failed, there was no way for the hosts to see the system vm template.

[weizhouapache]

The team noticed some errors during deployment regarding NFS SR being unable to mount on the xcp-ng hosts. So if nfs mount failed, there was no way for the hosts to see the system vm template.

correct.
have you fixed the nfs issue ? @alsko-icom

[alsko-icom]

Excuse my late reply, somehow i missed the notification of your update! Is there a problem that we have all 4 xcp-ng hosts connecting to the nfs? Not sure how to tackle. The NFS is not seen to be provisioned / mounted to the entire server pool of 4 but just to 1 server. So if the mount tries to switch to another server, the new server does not know it. Maybe we should try another approach? Also, when adding servers to the xcp-ng server pool, should they be added to the pool through cloudstack by being previously standalone or is it ok to be already added to the pool through an XO environment and the cloudstack will accept them in?

[DaanHoogland]

No @alsko-icom , unless there are some setting on the nfs server preventing it, all servers should be able to share the NFS storage.

[alsko-icom]

ok we will review the nfs behavior a bit more. Regarding the prep, should we add the xcp-ng hosts to the pool through cloudstack or through xen orchestrator at a lower level is fine and will not bother cloudstack?

[DaanHoogland]

@alsko-icom , I think you can add it to the pool and add the pool-master. please check though.