Suse Linux loses SAN connectivity

[Kassim Ismail]

Hi All,

anyone come across a problem before where a blade with Suse/Redhat linux installed loses connectivity when one of the HBA is disconnected? we have a situation where we have a blade with Suse linux 11 installed (SAN boot) & we configured 2 vHBA's (one on each Fabric) plus Powerpath is installed & configured correctly (as advised by SAN vendor).

we "disabled" one of the vHBA's & immediately we lost connectitity even though the 2nd vHBA was still "connected". as soon as we re-enabled the vHBA, it started working. we then tested by disabling the 2nd vHBA & left the 1st vHBA running & it still lost connectivity.

we will be looking at the powerpath logs, but i thought i would ask the forum to see if anyone can give some pointers,

any suggestions?

Thanks

Kassim

[Robert Burns]

Kassim,

How exactly are you "disabling" one of the vHBAs?

Robert

[Kassim Ismail]

Hi Robert,

i "disabling" the vHBA's by changing its VSAN to default from the correctly configured one.

Kassim

[Robert Burns]

Good deal - thats the preferred method.

Do you notice the issue with powerpath not installed?

Rob

[Kassim Ismail]

Robert,

have not tested without powerpath but i guess it will be the same.

reason for saying the above is that we found out the problem yesterday when SAN vendor was carrying out a FLARE code upgrade on the SAN & when SPA rebooted, our Suse server fell over. all other ESXi hosts continued working.

we checked the powerpath install & found that it was installed but was not registered. below was the output when querying the powerpath install:

"

# rpm -qa EMCpower*

EMCpower.LINUX-5.3.2.00.00-029

# powermt check_registration

There are no license keys now registered.

# powermt display dev=all

PowerPath modules are not loaded.

Ensure that the modules are loaded and try again.

"

so i guess this "might" equate to powerpath not been installed as its un-usable.

we did get it registered & configured & asked SAN vendor to confirm the config & they were happy with it.

Kassim

[Robert Burns]

If other hosts stayed (ESX) then this is not likely to be a UCS issue, but rather an OS/driver problem with SUSE.  I'll do some digging to see if we've seen any similar issues with SUSE 11.

Robert

[Kassim Ismail]

Thanks Robert.