100+ days to Resync on a new Storage pool

[wholybee]

I am installing Starwind SAN Fee on Proxmox, version 1.7.794.7576. Following this guide(from the download email):
https://www.starwindsoftware.com/resour ... ng-web-ui/

Configuration is Proxmox 9 on Dell VxRail 570
Two Starwind nodes with a witness
Three 10GBe networks for Management Sync and Data.
iperf3 shows networks working fine at 10Gbps.

I started with a test setup using single drives (what were the cache drives when these machines were running VMWARE VSAN). That worked fine, mounted the lun in Proxmox and everything tested ok.

Now trying to configure the other drives (6 per node) as a RAID1, and it gets stuck with a state of "Resync 108 d ..." I have deleted everything several times and can't get past it. It starts out as about 5 hours, and then climbs to over 100 days. ChatGPT suggested I create at 2GB volume and it would sync, that didn't work. Then it suggested I install version 1.8, which doesn't seem to exist, then it said this was a known problem with the Linux VM and I should install the Windows version instead. ChatGPT obviously can't fix this.

Suggestions?

[yaroslav (staff)]

If you are referring to MDADM resync, the issue is on the disk side (namely, discard support).
Try a parity RAID.

[wholybee]

I am referring to what is displayed in the management portal when I select the storage pool. I don't know if that is MDADM or not.
Also, I forgot to mention that I am passing the controller to the VM via PCIe passthrough. So I don't think discard support is applicable. Does the VM Appliance have any tools for checking or notifying physical disk errors? This was all working fine under VMware so I don't expect there is an issue, but if there is, I would need a way to check the status. Does Starwind have a tool for managing disks, notifying of a failed disk, r/w errors, or SMART status?

Results are the same with both RAID1 and RAID5.

[yaroslav (staff)]

I guess the controller is in HBA mode (i.e., no physical RAID). What you are trying to create is RAID in CVM. That's an MDADM. What you can do is try creating MDADM with CLI and use --assume-clean flag. That will skip the sync (not recommended way).
From my experience, longer RAID resyncs often point to the disks acting "funny" under discard commands.

[kumarbr]

A very long resync usually points to disk or RAID issues, so checking drive health or the RAID setup might help speed it up.kolkata fatafat result

[emmascott63]

Even with PCIe passthrough, discard/TRIM can still be relevant depending on how the controller presents the disks to the guest. I’ve seen cases where Linux MDADM resync slows to a crawl due to queue depth limits, NCQ quirks, or firmware that was “happy” under VMware but behaves differently under KVM.

[yaroslav (staff)]

You can't disable discard for every RAID levels; only parity RAIDs.
What you can do, though, is use --assume-clean flag. It skips initialization. Not ideal, but it helps.

[VirajIrani]

A prolonged resync often signals disk or RAID trouble, so checking drive health, firmware, and array settings can help.

[yaroslav (staff)]

That's an MDADM.

[kolkataau]

Great discussion here!

[Sharma]

This topic very interesting!

[khan76]

thanks for this information, this me a lots. Best Regards

[yaroslav (staff)]

You are always welcome.