HPE 9000 and HPE e3000 Servers
cancel
Showing results for 
Search instead for 
Did you mean: 

rp8400 I/O power issue

 
SOLVED
Go to solution
John Payne_2
Honored Contributor

rp8400 I/O power issue

I have an rp8400 that we powered down a few weeks ago to install memory. (Disconnection of power cables.) On replacing the power, one of the 2 I/O chassis's did not have any power to it.

When the machine went to mount the VG's, the machine panic'ed. (We have an EVA, with 2 HBA's here, one in each side of the machine.) After getting the power back to the I/O chassis, the machine booted normally.

Now for my question: Does a loss of power to 1 of the 2 I/O chassises bring a machine down. (If you pull the I/O power supply out, does the machine crash?) I would imagine the machine will keep running, SecurePath will failover, and things will be ok. I don't actually have a spare rp8400 available to test this with by pulling an I/O power supply. (Does anyone?)

Did my machine initially crash because the OS and Secure Path were not all the way up and the machine felt like the paths were missing?

We didn't fully resolve the power problem. (Read "The CE left without swapping any parts.") I am trying to get a finger on what our actual risk might be...

Thanks

John
Spoon!!!!
1 REPLY 1
Iain Ashley
Trusted Contributor
Solution

Re: rp8400 I/O power issue

Hi John,
The box will probably HPMC if you lose an entire IO chassis, mainly because the SBA for your core IO is present on the IO chassis backplane. If you lose the right one, you may be lucky but the RP8400 is not designed to be that fault tolerant. See this thread:

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=273447

The system probably paniced on boot because the autofile probably did not specify the "-lq" option. LVM can only activate a volume group if it does not have quorum, that is more than 50% of it's PVs are available. This means that if you have your root disk mirrored across controllers it will lose half of it's disks and therefore not have quorum. To rectify this, you need to either update your autofile or stop the boot at BCH and boot manually with the -lq option.

To update your autofile run the following command on you primary and alternate boot disks:

mkboot -a "hpux -lq" /dev/rdsk/

-lq tells LVM to ignore quorum when activating VG00 - everybody with mirrored root disks should be using this option


As for the power problem, is it working now? Did you try powering it on manually from the mp with the pe command. If you powered on the system with the front panel switch, sometimes you need to manually power on the cells and IO from the MP.

Regards
Iain Ashley