topic HBA failover in Operating System - Linux

HBA failover

unre4l — Wed, 20 Jun 2007 17:03:46 GMT

I am trying to test the failover of the HBA )Hp-1242SR)running on RHEL4 to an EVA 5000. I am using the hp linux driver kit. It has one active path to the EVA. I had a file transfer running and disconnected the active port and it failed over fine to port2. But when port1 was plugged back in the transfer cotinued on port2.
Sansurfer shows all other luns failed over to the alternate path and then failed back to the preffered path when the link came back except for the lun to which the data was being written to. That still shows as using alternate and is not failing back. Is there anything that needs to be done to force this? Reboot? Or is there any there command to do this?

Re: HBA failover

unre4l — Wed, 20 Jun 2007 18:29:55 GMT

I also see this in the var log messages just for the lun that was failoved over but did not fail back. The other luns show up fine.

qla2x00: no more failovers for request - pid= 564
SCSI device sdd: 849346560 512-byte hdwr sectors (434865 MB)
qla2x00: no more failovers for request - pid= 568
sdd: asking for cache data failed
sdd: assuming drive cache: write through
qla2x00: no more failovers for request - pid= 569
SCSI device sdd: 849346560 512-byte hdwr sectors (434865 MB)
qla2x00_combine_by_lunid: Found a controller path 0x0 - lun 0
qla2x00: no more failovers for request - pid= 573
sdd: asking for cache data failed
sdd: assuming drive cache: write through
sdd:<6>qla2x00: no more failovers for request - pid= 574
Device sdd not ready.
end_request: I/O error, dev sdd, sector 0
Buffer I/O error on device sdd, logical block 0
qla2x00: no more failovers for request - pid= 575
Device sdd not ready.
end_request: I/O error, dev sdd, sector 0
Buffer I/O error on device sdd, logical block 0
qla2x00: no more failovers for request - pid= 576
Device sdd not ready.
end_request: I/O error, dev sdd, sector 0
Buffer I/O error on device sdd, logical block 0
unable to read partition table

I am using the 2.6.9.55 hugemem kernel and the latest 8.01.06 version of the linux driver kit.

Re: HBA failover

Court Campbell — Thu, 21 Jun 2007 13:53:48 GMT

dumb question, for the virtual disk, what is the preferred presentation path?

Re: HBA failover

unre4l — Thu, 21 Jun 2007 19:12:52 GMT

Prefferd path is A. Mode is failover only. The strange thing is it only happens for one of the 4 luns presented. I have tried different driver versions too. It loads sdc fine but when it comes to sdd it errors out and then loads sde and sdf fine. I can do an fdisk on sdd once the system comes up and i repartitioned it and formatted it but still it errors when i restart the system. Here is the log when booting up.
Any help is appreciated.

Vendor: COMPAQ Model: HSV110 (C)COMPAQ Rev: 3028
Type: RAID ANSI SCSI revision: 02
qla2400 0000:0a:00.0: scsi(3:0:0:0): Enabled tagged queuing, queue depth 16.
Vendor: COMPAQ Model: HSV110 (C)COMPAQ Rev: 3028
Type: Direct-Access ANSI SCSI revision: 02
qla2400 0000:0a:00.0: scsi(3:0:0:1): Enabled tagged queuing, queue depth 16.
SCSI device sdc: 83886080 512-byte hdwr sectors (42950 MB)
SCSI device sdc: drive cache: write through
SCSI device sdc: 83886080 512-byte hdwr sectors (42950 MB)
SCSI device sdc: drive cache: write through sdc: sdc1
Attached scsi disk sdc at scsi3, channel 0, id 0, lun 1

Vendor: COMPAQ Model: HSV110 (C)COMPAQ Rev: 3028
Type: Direct-Access ANSI SCSI revision: 02
qla2400 0000:0a:00.0: scsi(3:0:0:2): Enabled tagged queuing, queue depth 16.
qla2x00: no more failovers for request - pid= 564
SCSI device sdd: 849346560 512-byte hdwr sectors (434865 MB)
qla2x00: no more failovers for request - pid= 568
sdd: asking for cache data failed
sdd: assuming drive cache: write through
qla2x00: no more failovers for request - pid= 569
SCSI device sdd: 849346560 512-byte hdwr sectors (434865 MB)
qla2x00_combine_by_lunid: Found a controller path 0x0 - lun 0
qla2x00: no more failovers for request - pid= 573
asking for cache data failed
sdd: assuming drive cache: write through sdd:<6>qla2x00: no more failovers for request - pid= 574
Device sdd not ready.
rc: Starting lm_sensors: succeeded Jun 21 18:43:13 bsdb kernel: end_request: I/O error, dev sdd, sector 0
Buffer I/O error on device sdd, logical block 0
qla2x00: no more failovers for request - pid= 575
Device sdd not ready.
end_request: I/O error, dev sdd, sector 0
Buffer I/O error on device sdd, logical block 0
qla2x00: no more failovers for request - pid= 576
Device sdd not ready.
end_request: I/O error, dev sdd, sector 0
Buffer I/O error on device sdd, logical block 0
unable to read partition table
Attached scsi disk sdd at scsi3, channel 0, id 0, lun 2

Re: HBA failover

skt_skt — Sun, 24 Jun 2007 09:08:37 GMT

The fail over to port2 is auotmatic .Fail back to port1 had not happened as the disk is in use and it may not be automatic

Could you unplug port2 and check if it is fail over to port1.If this functions well, you are safe interms of data integrity.

Re: HBA failover

Kumar Raman — Tue, 26 Jun 2007 10:43:58 GMT

Yes when i unplug port 2 it does fail over to port 1 and that works well. It's just that when it is failed over to port 2 and i reboot the luns on port 2 do not show up during boot up ad the system scans for luns only on port1.

Re: HBA failover

IT_2007 — Wed, 27 Jun 2007 13:43:17 GMT

what is vgdisplay shows at that time? which path shows as primary and alternate? If it fails over then vgdisplay should show port1 as alternate path.

No need to reboot server. since your file transfer continuing on port2, it won't fail back to port1 even if connect port1 back. But if you want to test it then during file transfer, unplug port2 and see if it fails back to port1.

Re: HBA failover

unre4l — Sat, 30 Jun 2007 13:32:58 GMT

Was a port error on the switch, changing to different ports on the switches fixed the errors.