Alerts not triggered by removing a drive.

[killervette]

I dont know if this is a bug or not, im very new to freenas. I have a ZFS Raidz1 with 3 drives. I physically disconnect a sata cable to remove the drive. The freenas webpage does not show any alerts. Viewing the drives in the area shows a drive as "unknown". The terminal screen on the server shows that it lost the ata device.

Is this normal? The array shows as healthy. It took a reboot to show the alert

 

[SoftDux-Rudi]

This is indeed actually a bug as it's not normal.

 

[jgreco]

I can confirm that this seems to be a problem. I pulled one drive out of a mirror. zpool status shows the state as ONLINE for the pool, but has a status note about an unrecoverable error. Oddly, the pulled drive shows as "ONLINE" as well, which is quite the amazing feat of wireless communications.

 

[killervette]

This bug really keeps me from switching my current setup to freenas. Without some sort of notification of a dead drive, I cant trust it. I dont want to have to restart the server every couple days just to see if my drive is bad. any eta on this?

 

[jgreco]

It appears that the situation is eventually detected, possibly by a scrub, but I agree that this is undesirable. There was some more discussion of it on the bug tracker, I don't have the ticket number handy.

 

[gcooper]

It appears that the situation is eventually detected, possibly by a scrub, but I agree that this is undesirable. There was some more discussion of it on the bug tracker, I don't have the ticket number handy.

I think this is what you were looking for: http://support.freenas.org/ticket/1066 . Either a zfs scrub or enough drive timeouts will cause the phantom drive to drop out of the zpool, but it could take a few days, depending on the machine usage.

Please note that not all chipsets support hotswap (in particular I know that older chipsets, pre-ICH7, couldn't do it).

What messages are you seeing on the console?

Also, I just had an epiphany on how this can be solved (not really elegantly, but it will do). I was reminded of a "hack" that someone at my old employer used to detect failed disks. I thought it was unnecessary, but now I understand why it was put in place.