"FC Fabric changed" on VCM, and "LOOP DOWN" on all BL460c in c7000

[Norisuke]

Hi,

We have a HP BladeSystem c7000 Enclosure with seven BL460c G1, two Onboard Administrator(ver. 2.04), four Virtual Connect Ethernet modules, and two Virtual Connect 4Gb FC modules.
Each BL460c has one QLogic QMH2462 4Gb FC HBA Mezzanine card, and 2 ports NIC Mezzanine card.
Each BL460c don't have internal hard disk, but boot RHEL4U5(x86_64) OS from SAN Storage HP StorageWorks EVA6100.

When I assigned one standby BL460c to the standby Server Profile of VCM(ver. 1.21),
the message "FC Fabric changed" was output on System Log of VCM,
and the message "kernel: qla2400 0000:10:00.0: LOOP DOWN detected" was output on system log of all BL460c in c7000.
Unfortunately all filesystems of LinuxOS on one BL460c changed to Read-Only, and the LinuxOS could'nt write logs in system log for a while.

My doubt is a change of the one Server Profile affects bad influences on all disk I/O of all blades in enclosure using Virtual Connect.

Can I change settings of Server Profile when blades is running ?
Does the message "FC Fabric changed" mean "Fabric status of all blades changed." ?

Thanks.

[HEM_2 · Accepted solution]

The FC Fabric changed is probably a general message indicating that the new HBA ports were recognized and/or being logged in to the fabric.

A single Server Profile change should not affect the other server profiles. The "LOOP DOWN" message doesn't sound good at all.

There was an issue in version 1.21 where certain changes (inserting a server) could cause configuration issues. This was fixed in VC FW 1.24. This might be what you are encountering - the insertion of a new blade may be causing issues for the VC-FC module configuration and causing the HBAs to be logged out of the fabric (this is just a guess though). Also there are some definite VC-FC fixes in VC firmware 1.24.

Please review the release notes here (pay special attention to page 15):
http://h20000.www2.hp.com/bc/docs/support/SupportManual/c01415160/c01415160.pdf?jumpid=reg_R1002_USEN

And it is a good idea to get the OA up to 2.20 and ILO's up to 1.50 to keep dependencies in check.

[Norisuke]

Thanks for your precious comment.
I think your guess isn't so far off the mark.
I'll try to upgrade VC FW, and check if recur or not.

For your information,
I found that latest HP StorageWorks Fabric OS of SAN Switch fix relevant bug.


http://h20000.www2.hp.com/bizsupport/TechSupport/SoftwareDescription.jsp?lang=en&cc=us&prodTypeId=12169&prodSeriesId=1143930&swItem=co-48535-1&prodNameId=1143931&swEnvOID=54&swLang=8&taskId=135&mode=4&idx=1

Traffic path between host and target not established on HP 4Gb VC-FC Module for c-Class BladeSystem due to PLOGI ACC frame not being routed to the destination virtual port when host and target are on the same NPIV port.
Fixed per Fabric OS 5.21b



http://h20000.www2.hp.com/bizsupport/TechSupport/SoftwareDescription.jsp?lang=en&cc=us&prodTypeId=12169&prodSeriesId=1143930&swItem=co-54655-1&prodNameId=1143931&swEnvOID=54&swLang=8&taskId=135&mode=4&idx=1

Change made in 5.2.0 to limit the number of FDISCs per unit of time, caused timeouts of NPIV logins in some BladeServer configurations with VC-FC.
Fixed per Fabric OS 5.3.0.


I'm looking into upgrade for Fabric OS, too.

Thanks.