HPVM Guest Insufficient Swap Resource

[Jason Shannon]

I have an RX4640 machine running Integrity VM 3.5 with 64GB of memory (just upgraded from 32GB). The OS is 11.23 Dec'07

The machine hosts 4 VM's.

To make full use of the memory upgrade, the OS was rebuilt and the swap increased from 32GB to 72GB.

I have increased the memory in the 4 guests to take into the account the hardware upgrade but receive the following:

[ni-cr-dev-vm4] vMP:CM> pc -on

System will be powered on.
Confirm? (Y/[N]): Y

-> System is being powered on.

Please wait for the guest start sequence to complete. Use of the attention
character can prevent the guest from running.
HPVM guest ni-cr-dev-vm4 configuration problems:
Warning 1: There are insufficient host swap resources for this guest.
These problems may prevent HPVM guest ni-cr-dev-vm4 from starting.
hpvmstart: Unable to continue.

-> Command failed.

[ni-cr-dev-vm4] vMP:CM> exit


vMP MAIN MENU

CO: Console
CM: Command Menu
CL: Console Log
SL: Show Event Logs
VM: Virtual Machine Menu
HE: Main Help Menu
X: Exit Connection

[ni-cr-dev-vm4] vMP> x
# swapinfo
Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 75497472 0 33554432 0% 0 - 1 /dev/vg00/lvol2
reserve - 17317472 -17317472
# hpvmstatus
[Virtual Machines]
Virtual Machine Name VM # OS Type State #VCPUs #Devs #Nets Memory Runsysid
==================== ===== ======= ========= ====== ===== ===== ======= ========
ni-cr-dev-vm1 1 HPUX On (OS) 2 4 1 16 GB 0
ni-cr-dev-vm2 2 HPUX Off 1 2 1 8 GB 0
ni-cr-dev-vm3 3 HPUX Off 1 2 1 8 GB 0
ni-cr-dev-vm4 4 HPUX Off 2 4 1 16 GB 0
#
#

Any suggestions?

[Mike Shilladay]

Hi Jason,

I think it maybe somehting as simple as you have increased the VM Guest memory beyond the size of the VM guest's primary swap. Check the swap space on the guests and make sure it is not lower than physically allocated RAM.

Mike.

[Don Morris_1 · Accepted solution]

First -- that "swapinfo -atm" is easier to read.

Second -- your device swap line, if I read it right -- implies to me that you need to increase swchunk (which isn't surprising, since the default value means the system goes up to 32Gb of swap out of the box). You should have seen a message on boot to this effect and with a recommended value for swchunk, by the way. (By my math, you want 4608). "man swchunk" for more details.

Long and the short of it -- the Host knows the device has another 50Gb, but can't create the metadata to manage it due to swchunk settings. So you really only still have 32Gb of swap. Over 16Gb is reserved, so you can't reserve the next 16Gb you need for vm4 (I would suspect vm3 or vm2 would come up -- and that would be about it).

[melvyn burnard]

I agree with Don, look at your swapdevice:
dev 75497472 0 33554432 0% 0 - 1 /dev/vg00/lvol2
Only 32Gb is free!, Check your syslog.log, and/or /etc/rc.log, and I bet there is amessage about tuning maxswapchumks.
You will need to increase that paraqmeter, and then all should be fine.

[Steven E. Protter]

Shalom,

The same memory rules apply to a vm machine as a regular machine.

swap must be at minimum 50% of physical RAM to insure stability.

I would set up swap to 100% of physical ram in any vm machine and not do a machine with less than 2 GB of RAM.

SEP

[AIMC]

Hi Jason,

You can try increasing the swchunk tunable to 4068. To see the current value run
kctune -q swchunk

If this doesn't solve the issue you may need to reduce the amount of ram on each back to its original value, boot the VM's & increase the amount of swap on the guests to reflect the changes you will/want to make.

PS. Not on mail today but keep me updated.

Regards,

Aidan

[Jason Shannon]

Thanks to all, particularly Aidan and Don who were exactly right with their suggested fix.