LUNs are not coming automatically under veritas after reboot

[Pranav_Tyagi]

Hi Community,

I have installed SFCFSHA cluster solution on RHEL 1.6, storage is connected through Fiber Channel Over Ethernet (FCoE) interace and dmp is used as multipathing solution.

The issue is while i reboot the system, I/O fecning got failed, disks are not coming automatically under veritas control. When i scan the disks and configure IO fecning, cluster starts working.

But i am unable to understand why disks are not coming under veritas control automatically.

Error -

MMTE03:~# tail -100f /var/VRTSvcs/log/engine_A.log
2015/07/08 20:59:55 VCS CRITICAL V-16-1-10037 VxFEN driver not configured. Retrying...
2015/07/08 21:00:00 VCS CRITICAL V-16-1-10037 VxFEN driver not configured. Retrying...
2015/07/08 21:00:05 VCS CRITICAL V-16-1-10037 VxFEN driver not configured. Retrying...
2015/07/08 21:00:10 VCS CRITICAL V-16-1-10037 VxFEN driver not configured. Retrying...
2015/07/08 21:00:15 VCS CRITICAL V-16-1-10037 VxFEN driver not configured. Retrying...
2015/07/08 21:00:20 VCS CRITICAL V-16-1-10037 VxFEN driver not configured. Retrying...
2015/07/08 21:00:25 VCS CRITICAL V-16-1-10037 VxFEN driver not configured. Retrying...
2015/07/08 21:00:30 VCS CRITICAL V-16-1-10037 VxFEN driver not configured. Retrying...
2015/07/08 21:00:35 VCS CRITICAL V-16-1-10037 VxFEN driver not configured. Retrying...

Thanks

Pranav

[Marianne]

Sure about RHEL 1.6 version? Which SFCFSHA version?

Have you confirmed HCL/array compatibility before installing SFCFSHA ?
Compatibility lists can be found over here: https://sort.symantec.com/documents  
HCL for SF 6.1 and 6.2: http://www.symantec.com/docs/TECH211575  
(I can see only a small amount of Array models supported with FCoE)

Also see the Hardware TN for array-specific configuration requirements:  http://www.symantec.com/docs/TECH47728 

(Use the 'Attachments' link in both URLs to open the pdf's.)

You need to confirm at OS level that disks/luns are visible before SF/fencing/HA starts up.

If there are any delays at OS level at boot time, you need will need to modify startup scripts to scan device earlier in the boot process and check that luns are visible before Veritas startup scripts are initiated.

[mikebounds · Accepted solution]

Are you saying that if you run a "vxdisk list" after booting you don't see disks and so you have to scan disks to see disks and then start fencing - is so then I would guess your FCoE drivers are starting after vxvm, so you will need to move starting of FCoE drivers to earlier on in the boot process.  If I have misunderstood your issue, then please describe what what you are missing after boot up and what you run to fix.

Mike

[Pranav_Tyagi]

Thanks Mike, your guss is absolutly correct, VXVM services are running before fcoe.

I have changed the boot priority of the VxVM and fcoe services, but still VxVM services are running before fcoe. Could you suggest something for this.

#chkconfig --list|grep "1:on"
fcoe            0:off   1:on    2:on    3:on    4:on    5:on    6:off
lldpad          0:off   1:on    2:on    3:on    4:on    5:on    6:off

Thanks

Pranav

[Pranav_Tyagi]

Thanks Marianne, i have wrongly mentioned OS version, its 6.5 and SFCFSHA version is 6.2.

I am using VNX 7600 for storage. I have checked boot logs and found that VxVM services are starting before fcoe services, and i tried to change the priority of these services from Linux chkconfig utility, but still VxVM is running before fcoe. 

[mikebounds]

So do you have fcoe starting in RC2 and if so what "S" number is it.

I don't see why this doesn't work, but the other alternative would be to create a new RC script in RC3 which starts before fencing which does a "vxdctl enable" (or vxdisk scandisks) 

Mike

[m_karampasis]

Hello Pranav,

We had the same issue with fencing disks attached FCoE and symantec suggested us to do the following changes:

(Case # 08995143)

1. Add the rdloaddriver=bnx2fc to /boot/grub/grub.conf.

2. Add DEV_DISCOVER_DELAY=120  in /etc/vx/vxvm-startup.

3. Reboot the nodes.

Now the issue is solved.

BR,

[Pranav_Tyagi]

Thanks Mike, boot-priority was not changed correctly earlier, after correcting it and adding a timer of 120sec in fcoe start-up script, it worked.

Thanks m.karampasis.