We encountered this weird ouput on the p795’s propoerties on the HMC, I never saw that before :

After a few research, I managed to find the partitions which had some pending values (I assumed this was the problem):

 hscroot@HMC:~> lshwres -r proc -m Managed_system --level lpar -Flpar_name,pend_proc_units,curr_proc_units
 lpar1,0.2,0.4
 lpar2,0.2,0.4

As you can see, the current EC is 0.4 and the pending (target) value is 0.2 .
The profiles are correctly configured, as if someone tried to downsize the EC, changed the profile, but did’nt (want to) see that the DLPAR operation had failed and left it that way:

 hscroot@HMC:~> lssyscfg -r prof -m Managed_system -Fname,desired_proc_units
lpar1,0.2
lpar2,0.2

So after a little digging, we found that command could do the trick : rsthwres

Description

The rsthwres command will restore the hardware resource configuration following a failure in a dynamic LPAR reconfiguration. When a DLPAR operation fails to complete successfully, the hardware resource assignment in NVRAM of the managed system and in the AIX OS running in the partition become inconsistent, and this command fixes the condition:
rsthwres -m « managed-system » [ -p « partition-name » ] -r [cpu | mem | slot ] [ -u < processor id > ] [[ -i < drawer id > -s < slot id > ] | -l « physical location code » ] [ —help ]
where:

• -m – the name of the managed system where the hardware resource is configured. This name must be enclosed in double quotes and in the form mmm*ssss, where mmm is the machine type and ssss the serial number of the managed system.

• -p – the user-defined name of the partition on which to perform the restore. This argument must be enclosed in double quotes.

• -r – the hardware resource type to restore. Possible values are cpu (processor), mem (memory), or slot.

• -u – the processor ID of the processor to restore.

• -i – the machine type and serial number of an I/O slot’s scoping drawer. This argument should not be specified if the resource type is not slot.

• -s – the slot number of a PCI slot to restore.

• -l – the physical location code of the I/O slot. This argument cannot be specified with the -i and -s flags.

• -help – prints a help message.

 hscroot@HMC:~> lshwres -r proc -m Managed_system --level lpar -Flpar_name,pend_proc_units,curr_proc_units
lpar1,0.2,0.4
lpar2,0.2,0.4
lpar3,0.2,0.2
hscroot@HMC:~> rsthwres -r proc -m Managed_system -p lpar1
hscroot@HMC:~> lshwres -r proc -m Managed_system --level lpar -Flpar_name,pend_proc_units,curr_proc_units
lpar1,0.4,0.4
lpar2,0.2,0.4
lpar3,0.2,0.2
hscroot@HMC:~> rsthwres -r proc -m Managed_system -p lpar2
hscroot@HMC:~> lshwres -r proc -m Managed_system --level lpar -Flpar_name,pend_proc_units,curr_proc_units
lpar1,0.4,0.4
lpar2,0.4,0.4
lpar3,0.2,0.2

We can see now that the pend_proc_units are back to their original value (before the DLPAR operation), which is the target value (in our case, the same as curr_proc_units)!

The properties are back to normal too :

Tip : 1-liner to get the pending lpars

You need to remote execute this, because of awk:

# ssh hscroot@HMC lshwres -r proc -m Managed_system --level lpar -Flpar_name,pend_proc_units,curr_proc_units | awk -F, ' $2==$3 {print $1}'
lpar1,0.2,0.4
lpar2,0.2,0.4

 Links

rsthwres manpage

http://www.coolcommands.com/index.php?option=com_cccat&task=display&id=216

Share Button
HMC : Do you have any « pending » values on your power server ?
Taggé sur :            

Laisser un commentaire