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

SD32A (PA8800 SuperDome) - Panics due to Memory Region Anomaly

 
Alzhy
Honored Contributor

SD32A (PA8800 SuperDome) - Panics due to Memory Region Anomaly

A yearling SuperDome 32000 (5 nPars). Two weeks ago, we had an nPar that was rebooting daily on Panics (panic String: SpinLock DeadLock and WSIO error messages just bfore the panics). After proving that it's not a bad OS (by actually booting the OS on a known good nPar) - HP decided to replace an entire I/O chassis with all the PCI cards on it. Case solved.

Of particular interest are the following memory registers which HP says should have non-zero values:

kfh_lock.sl_lock
kfh_lock.sl_name_ptr
kfh_kap

We used the following script to look at this structure:

q4 ./vmunix /dev/mem <
Hakuna Matata.
1 REPLY 1
Jeff Schussele
Honored Contributor

Re: SD32A (PA8800 SuperDome) - Panics due to Memory Region Anomaly

Hi Nelson,

On our 'Domes we see the following failures in order:

1) Power Supplies
2) I/O cages
3) I/O cards
4) Cell boards

But we rarely are not alerted by EMS and we've only paniced a couple of times compared to the dozens of replacements we've endured. So make sure your OnLine Diags are up to date.
I suspect that indeed an I/O cage or an I/O card or the culprits in both instances.

BTW - we're moving away from 'Domes due to the failure rates we experience as well as financial reasons. The lease/purchase costs as well as maintenance costs are becoming prohibitive. We have found that rp8xxx systems are just as fast & if we need more I/O slots we use the System Expansion Units.

My 2 cents,
Jeff
PERSEVERANCE -- Remember, whatever does not kill you only makes you stronger!