Boot from SAN with Quadstor not working

[Timo]

Good morning,

we made a new atempt with Quadstor on CentOS7 with latest 3.2.11.

All worked well so far. Also the "fencing" with fcconfig that only the Client is able

to see the FC LUN it should see.

The vDISK are without any parameters (no D,E,C).

In the client we made the QLA-Adapter bootable and assigned a LUN.

When booting the client with Ubuntu oder CentOS CD for Install, all worked well.

They can see the disk and install also worked fine.

Yet when we try to reboot, nothing happens any more, except this message in /var/log/messages on the quadstor

Mar  3 09:43:32 XXX kernel: rport-1:0-33: blocked FC remote port time out: removing rport

Mar  3 09:43:32 XXX kernel: rport-7:0-33: blocked FC remote port time out: removing rport

And then nothing. It does not boot. 

Any advice? What are we doing wrong?

Client and Server are both Dell R710 with QLA2562 on two fc fabrics with Qlogic SANboxes.

Any help would be welcome.

Thx!

[QUADStor Support]

The kernel messages indicate that the initiator port is no longer
active (no further FC session from that side). So it seems that after
the reboot the initiator port is probably not connecting back to the
target.

Does fastutil list the LUN after the reboot and is it marked as the boot disk ?

> --
> You received this message because you are subscribed to the Google Groups
> "QUADStor Storage Virtualization" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to quadstor-vir...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Timo]

Hi,

Am Freitag, 3. März 2017 20:00:11 UTC+1 schrieb quadstor:

The kernel messages indicate that the initiator port is no longer
active (no further FC session from that side). So it seems that after
the reboot the initiator port is probably not connecting back to the
target.

Does fastutil list the LUN after the reboot and is it marked as the boot disk ? 

So bootet into fastutil. And indeed, it did not list the LUNs or anything related to Quadstor any more.

Why?

What is wrong?

How can we fix that?

We have another setup for a test with targetcli, there it seems to work.

The hosts an FC are identically configured. 

Except the target comes from targetcli...

Thx!

Best,

Timo

[QUADStor Support]

Anything on the fcconfig side which might be causing this. Are the
vdisks emulating 512 byte sectors ?

Try clearing all fcconfig rules, restart the quadstor system, then try
a restart at the initiator side and let us know. There is a pending FC
update which might be available next week, but boot from SAN should
have worked with the current version also.

[Timo]

Hi,

thx for your help. We figured out, what the problem is:

cat /sys/module/qla2xxx/parameters/qlini_mode

enabled 

Which means, your qlainst module doesn't disable the Initiator-Mode.

So we created a file:

cat /etc/modprobe.d/qla2xxx.conf

With input:

options qla2xxx qlini_mode="disabled"

Yet it did not work, after a reboot. The qla2xxx module didn`t use it.

Then we figured, the /quadstor/bin/builditf didn't recognize the option while building (for latest kernel of centos 7.3).

So we modified line 11 of your qlainst:

/sbin/modprobe -q qla2xxx 

with this slight change:

/sbin/modprobe -q qla2xxx  -C /etc/modprobe.d/qla2xxx.conf

Then we did the builditf again and after a reboot we could see:

cat /sys/module/qla2xxx/parameters/qlini_mode

disabled 

After all that, the other hosts are now able to boot from this LUN.

The rport errors at the quadstor host (var/log/messages) is now gone for good.

HTH,