Ouch - Upgrade Bluefin to Cobia, Major iSCSI Issues

[mevans336]

I just upgraded the latest Bluefin to the latest Cobia and am experiencing major iSCSI issues.

First, it deleted one of my file extents outright. That stings, but I have backups. (LUN 0)

Second, for my other LUN (LUN 1, different zfs array) the file is still there, but when I try to start the iSCSI service crashes and says it can't open the file, to run "dmesg". This of course causes the service not to start. The dmesg error is

Code:

dev_vdisk: ***ERROR***: Parameter t10_dev_id value missed for device ssd-pool

.

Looking in /etc/scst.conf, t10_dev_id doesn't have a value and actually has a huge block of whitespace characters after it.

Code:

HANDLER vdisk_fileio {
    DEVICE ssd-pool {
        filename /mnt/ssd_pool/ssd-iscsi.extent
        blocksize 512
        read_only 0
        usn
        naa_id 0x6589cfc0000000d762b3b13173278485
        prod_id "iSCSI Disk"
        rotational 0
        t10_vend_id TrueNAS
        t10_dev_id
    }

If I go back into the properties of the extent and add a random value to the "serial" the service will now start.

So, ummm, yeah. What a disaster.

 

[ABain]

@mevans336 I'm going to try and reproduce this today, can you tell me which version of Bluefin you were running (latest major release or nightly build).
Please can you file a bug report : https://ixsystems.atlassian.net/jira/software/c/projects/NAS/issues
A link will appear when you create a ticket for a private debug upload.

 

[HoneyBadger]

Hi @mevans336

We've managed to reproduce the serial-number t10_dev_id issue and we're narrowing down a way to implement a fix for the release of Cobia. We really appreciate you bringing this to our attention.

Regarding the sudden disappearance of your LUN0 extent - did it actually delete the backing file, or just the target-extent linking in the scst.conf?