Ubuntu 20.04 Template Problem

[rastopato]

problem

Deployment of VMs from Ubuntu 20.04.06 LTS stopped working after upgrading ACS to 4.21 and VMware to 8.0.3. Every VM deployed from this template is stuck at boot with corrupted ext4 filesystem errors.

I tried every possible combination of preparing new ubuntu template, disk adapters, bios/uefi, etc, etc, nothing worked.
Ubuntu 22.04 and 24.04 has no problem.

I tried it on environment where are previously deployed Vms from Ubuntu 20.04 template without issues. Now, after upgrade, it is the same error.
I got this error in all my production and testing environments.

Finally i tried it on environment on ACS 4.20.1 and VMware 8.0.0 and there is no problem with Ubuntu 20.04.

management-server.log.zip

versions

ACS: 4.21
VMware vCenter 8.0.3
Vmware ESXi 8.0.3
Ubuntu 20.04.06 LTS
Primary storage is VMFS datastores on SAN storage

The steps to reproduce the bug

1. Create Template with Ubuntu 20.04.06 LTS
2. Deploy new VM from it
3. It should be stack at the boot with filesystem corruption error.
4. If you not prepare any /etc/cloud/cloud.cfg.d/ files, it maybe start, but you cant even write username/password because of corrupted filesytem

What to do about it?

Please resolve it, or suggest workaround.

[weizhouapache]

You may refer to #7572 and #9625

[rastopato]

i'm not sure, what you are refering to. esx 8U3 is supported from version 4.20.1

[weizhouapache]

i'm not sure, what you are refering to. esx 8U3 is supported from version 4.20.1

if you go through the issue #7572, you might know that we face a similar issue with cloudstack systemvm template since VMware 8.0.1. The root cause is still unknown but we have addressed the issue by #9625. To be clear

• add fsck.mode=force fsck.repair=yes to boot args
• add depmod -a in init.sh

The same issue might impact user templates too. This is why I asked you to have a look.

[rastopato]

I assumed if these issues are closed and compatibility matrix is ok with ESX 8.0.3, it will be ok. Anyway, all other templates are ok with 8.0.3, except Ubuntu 20.04. I added fsck.mode, fsck.repair and depmod -a. I'm now little step closer, but it is still not working:

[weizhouapache]

I assumed if these issues are closed and compatibility matrix is ok with ESX 8.0.3, it will be ok. Anyway, all other templates are ok with 8.0.3, except Ubuntu 20.04. I added fsck.mode, fsck.repair and depmod -a. I'm now little step closer, but it is still not working:

You may need to run "depmod -a" when create the template.
CloudStack uses packer, it is one of the last steps in the process, see #11128

[rastopato]

I did, of course. Now I tried to change the way i ran it, and i ran it at the end, and it looks promising now. I will let you tomorrow.

[rastopato]

Well, no luck. i run depmod -a, set fsck in the grub. Fsck is running during booting process, fix a lot of things, but VM is obviously still corrupted:

[sureshanaparti]

@rastopato We fixed similar file system errors with the below options mentioned by @weizhouapache already, in this PR: #9625. Maybe, it's hitting different issue now. Can you try to create a VM in vCenter/ESXi directly using this template and check.

• add fsck.mode=force fsck.repair=yes to boot args
• add depmod -a in init.sh

[rastopato]

Hi guys, i've done a little bit more troubleshooting based on your comment and I found out some issues. When i prepared Template VM i export it from VMware using vmware ovftool.

When I import this ova to the target vmware environment, it is working with no problems.

When I import this ova to the ACS as template it is saved on NFS storage in separate directory with some files in it. There is whole .ova file with different name (3ff4f989-6751-3607-9fd1-63c1429da88d.ova) and most probably the same .ova but unpacked to vmdk, ovf, nvram and mf file.

When I try deploy OVF template directly on VMware from these unpacked 4 files, i got error: "Failed to deploy OVF package. Cause: A specified parameter was not correct: The checksum(s) from the provided manifest file do not match the content of file(s): ubuntu2004.ovf."

When I copy whole .ova file from NFS storage (for example 3ff4f989-6751-3607-9fd1-63c1429da88d.ova), unpack it manually and Deploy it as OVF directly on vmware, it is working

Then i tried to delete 4 files from directory on NFS and unpack there that .ova file, delete all templates from vCenter and also from primary storage. Then i tried deploy VM again from ACS, but still no success - corrupted filesystem.

I just realize that all Templates which were working on previous versions are now broken. I was able to repair some of them by simply regenerate and import new template (Ubuntu 22.04, 24.04, RHEL9.x), some of them are broken with corrupted filesystems also after regenerating it (Ubuntu 20.04, Rocky Linux 9.x, CentOS Stream 9)

[rastopato]

It is definitely corrupting filesystem when VMware Templates are creating. During the process of creating instance, multiple VMware templates are created by deploying .ovf template from NFS storage to Primary NTFS storage. This templates are already corrupted, when i try to deploy Vm from it directly in VMware.