S.M.A.R.T tests in Proxmox – best practice?

[Raid007]

Hi everyone,

today I just realized that Proxmox only displays the current SMART values of the disks in the GUI.
It seems that no scheduled SMART self-tests (neither short nor long) are actually executed by default.

• Is that correct, that Proxmox does not run any SMART tests on its own?
• ZFS scrubs are running regularly, but from my understanding it would still make sense to run SMART tests in addition, right?
• What is the current best practice for SMART testing on Proxmox? How often should a short test / long (extended) test be performed?
• How should these tests be scheduled?
• And is it possible to get notifications about SMART test results via the Proxmox notification system (same as for backup jobs, etc.)?

Thanks in advance for your insights!

 

[LnxBil]

Is that correct, that Proxmox does not run any SMART tests on its own?

Yes, neither does any system I know. You need I/O predicability, so running stuff in the background is a no-go for an enterprise.

ZFS scrubs are running regularly, but from my understanding it would still make sense to run SMART tests in addition, right?

A scrub is the real world metric you need. SMART is not that good at predicting failures.

I don't run SMART tests unless there is an error. ZFS scrub is more sensitive to problems in my experience and running SMART analysis after an scrub yielded an unrecoverable error on the disk, did not yield any abnormal tests. In general, you will have better results if you use better hardware, e.g. SAS instead of SATA. The enterprise firmware on them will report back problems much earlier than any SATA firmware would, so you will notice much quicker if a disk dies.