MSL 6000 Library shutdown causes Linux server to reboot

[parkarnoor]

Hi,

i have a very critical issue on production servers.

i have 4 servers with 2 each in cluster.

The problem is when i shutdown MSL 6000 Library by pressing power button in front - which is in another rack it caused these 2 servers to reboot.

MSL 6000 Library is in SAN. and not directly connected to servers.

what can be the problem??

what tools can be installed on the server to diagnose this problem

Regards,
parkar
UAE

[Mudgen]

HP MSL600 tape library? An EVA3000 or similar SAN? SCSI connected to the SAN or its embedded Windows Appliance Version management console? Or connected to the fiber network (in which case it IS connected to the servers)? Which 2 of the 4 servers reboot? Both in one cluster, or one in each?

If fiber connected, my guess would be some misconfiguration in failover, perhaps in the modprobe options for the server HBA drivers. If SCSI connected as described, it makes no sense that the servers would reboot.

Edit: have you checked /var/log/messages for anything logged just before the reboot?

[parkarnoor]

Hi,
i checked /var/log/messages.. there is nothing. it is immediately restarting the server.

In fact if we remove power cable from MSL library from the rear it is rebooting other 2 linux server also.

So it is rebooting all 4 servers not together.

i mean if i pressed power button in front of the library then it will restart two linux server in one cluster then another two servers in one cluster also

MSL is connected to fibre network. i will update you about the setup soon.

Regards,

[parkarnoor]

Hi,

we have total fibre based setup.

Linux server are connected to and from SAN switches and we have SAN-EVA400.

MSL library is also connected to SAN switches.



Regards,

[Mudgen]

HP servers as well? I'm guessing you're using the HP-branded QLogic cards, with their HP drivers, which do failover HP style. What does your modprobe.conf look like?
You say "cluster"--what clusterware? Are we talking about an OS-level cluster, or an Oracle RAC, or some other application layer cluster?

What you're seeing resembles Oracle "fencing" in some ways, FWIW. I'm still guessing that there's some misconfiguration of failover, heartbeat, or the like that causes the reboots. As if, perhaps, when the library drops out one or more servers think they've lost a LUN.