driver specific files need to be separated with a dot (.)

[chiraganand]

See this link:

restore.<SYSTEM_TM>: [Only for KVM drivers] If this script exists, the restore script will execute it right at the beginning to extract the checkpoint from the system datastore. For example, for the ceph system datastore the restore.ceph script is defined.

1. File: vmm/kvm/save_linstor_un should be renamed to vmm/kvm/save.linstor_un.
2. File: vmm/kvm/restore_linstor_un should be renamed to

[kvaps]

Hi, there were big discussion about this
https://forum.opennebula.io/t/shared-mode-for-the-ceph-based-system-datastires/7164/3?u=kvaps

Actually save_linstor_un and restore_linstor_un are not exactly same as save.linstor_un and restore.linstor_un, the main difference that save.linstor_un and restore_linstor_un handled by save and restore scripts of vmm driver remotely on the compute node. But we need to handle them as any other tm actions on the controller side, because it has an access to Linstor API.

There were no other option except overriding standard save and restore vmm actions by local ones, you're doing by updating the VM_MAD driver config:

 VM_MAD = [
     NAME           = "kvm",
-    ARGUMENTS      = "-t 15 -r 0 kvm",
+    ARGUMENTS      = "-t 15 -r 0 kvm -l save=save_linstor_un,restore=restore_linstor_un",
 ]

BTW there is an unfinished PR on upstream project OpenNebula/one#3273, which aims to add presave, postsave, prerestore and postrestore tm actions, which actually solves this issue.

[kvaps]

Unfortunately I can't accept your PR because it will break the driver logic by calling save.linstor_un and restore.linstor_un on the remote node side which will never be finished successfully.

[chiraganand]

Hi, there were big discussion about this
https://forum.opennebula.io/t/shared-mode-for-the-ceph-based-system-datastires/7164/3?u=kvaps

Actually save_linstor_un and restore_linstor_un are not exactly same as save.linstor_un and restore.linstor_un, the main difference that save.linstor_un and restore_linstor_un handled by save and restore scripts of vmm driver remotely on the compute node. But we need to handle them as any other tm actions on the controller side, because it has an access to Linstor API.

Thanks for the explanation, Andrei. This makes sense.

There were no other option except overriding standard save and restore vmm actions by local ones, you're doing by updating the VM_MAD driver config:

I have re-deployed linstor_un using this now and everything is working fine. I am guessing there was an issue with these files names on the earlier OpenNebula (5.6) which seems to have gone now. Thanks!