Pagefile read error at boot time

[Jansen_8]

Hi all,

Since some time I have the following problem on the bootmember of my OpenVMS cluster:
-all memembers run openVMS8.3 AXP
(patches up to date)
-one bootmember and 4 satelites sharing the
system disk of the bootmember
-All satelites have their page/swap files on
local disks.
-The bootmember has 2 sets of page/swap files:
Swap File Usage (8KB pages): Index Free Size
DISK$ALPHASYS:[SYS0.SYSEXE]SWAPFILE.SYS
1 56 56
DISK$POLKA2:[SYSEXE]SWAPFILE1.SYS;2
2 1448 1448

Total size of all swap files: 1504

Paging File Usage (8KB pages): Index Free Size
DISK$POLKA2:[SYSEXE]PAGEFILE1.SYS;2
253 16520 16632
DISK$ALPHASYS:[SYS0.SYSEXE]PAGEFILE.SYS
254 8 56

Total size of all paging files: 16688
Total committed paging file usage: 9516


Once in a while the bootmemeber hangs and becomes irresponsible. After reseting the machine it fails to boot giving an error about being unable to initialize the page file because of a failure on the configuration block
(the exact error moves to fast from the screen because of the dump that follows). The only way to get the system running again is booting from CD and creating a new pair of pagefile.sys/swapfile.sys on the system disk.This means that the whole cluster has to be taken down since mounting the system disk results in a "rebuild"
What can be the problem? I do not have much to go on because:
-No system dump is made since the machine only hangs and did not crash
-No (hardware) errors at all are logged.
-Appart from the pagefiles, no other files on the same hard-drive seem to be corrupted.
-Analyze/disk/repair does not help

Any idea how to solve this?

Jouk

[Hoff]

We really need that diagnostic and the stackdump.

Connect a serial console and capture it.

Post some details on the server and the storage, as well.

[Joseph Huber_1]

DISK$ALPHASYS:[SYS0.SYSEXE]PAGEFILE.SYS
254 >>> 8 56

Why do You create such an incredible small page file at all ? 56*8K = 448K bytes!

I don't wonder the system hangs (although I did not dig deeper...).

Delete the system disk pagefile if You need the space urgently, or create a much bigger one, then reboot.
(also the 133MB page file on disk$polka2 seems to be rather small: are You that short on diskspace ?).

[Jansen_8]

Next time it happens I'll try to capture the dump (can take some time)

The bootmember is a DPWS600a with 256MB RAM.

The reason for the very small page/swap files on the system disk is indeed lack of disk-space there. The pagefile1.sys and swapfile1.sys are on another local disk which has much more space. The idea is to have some small page/swap space to get it booting and add the second set during running of the systartup_vms.com when the second disk is booted. The sizes of this second set as suggested bu autogen and I do hardly ever see more usage of this space than 10%.
The hang occurs when both sets are active.

[Joseph Huber_1 · Accepted solution]

>>The idea is to have some small page/swap space to get it booting and add the second set during running of the systartup_vms.com

Well maybe the dump analysis reveals some reason other than the page-file too small
(Is the system also set to dump to the page-file ?)
, but I still believe no page file at boot (on the boot disk) is perfectly valid: the secondary one on the local disk should not be installed in systartup_vms (where it could be too late), but in sys$system:SYPAGSWPFILES.COM or satellite_page.com: they are executed long before systartup_vms from VMS$DEVICE_STARTUP.

[Hoff]

Missed that. That's almost certainly the problem. A corner case. Get rid of that page file. Or size it reasonably. Used 18 GB DEC/Compaq SCSI disks are cheap; you may pay more for the shipping than for the hardware.

[Steve Reece_3]

Jansen,

It's perfectly valid, as Joseph and Hoff have said, to not have a page or swap file on the system disk. The only time that you'd actually need one is if you booted minimum and had no other disks mounted, though this is likely to be an environment that doesn't need a swap or page file anyway!

Get rid of the files on the system disk and see what happens. It's likely to solve your problems.
I'm not sure whether it's been updated, but VMS certainly used to use a page file on the system disk in preference to one elsewhere so this is likely to cause you problems anyway having the itsy bitsy pagefile on the system disk and a bigger one elsewhere.

Steve

[Hoff]

Recent VMS doesn't particularly care about the pagefile disk location and will happily use pagefiles on other disks, but it does get cranky about a bogus pagefile configuration and it gets cranky (bad performance, weird wedges, pauses) when the system disk approaches capacity.

You need to move or delete files off this system disk or otherwise free up about 20% of the space), or you need to add another disk for pagefiles, or to offload the users and the non-system files, or you need to replace these disks with larger disks.

RAID-1 or better would be a good idea via HBVS or via a RAID controller, while you're adding disks to the environment. These older disks do die, and I'm guessing this is a pretty old and constrained configuration, and that it's likely vulnerable to a crash.

[John Gillings]

Jouk,

>The idea is to have some small page/swap
>space to get it booting and add the second
>set during running of the
>systartup_vms.com when the second disk is
>booted.

OpenVMS doesn't need a pagefile until you start running out of memory. Even with as little as 256MB of RAM, you can do a lot of work in memory. I'd be astonished if you needed a page file of any size before the startup process reached the point when it could mount a secondary disk and install a pagefile.

Having a tiny pagefile can be worse than having no pagefile at all (though V8.3 has greatly improved page and swap file management over older versions).

If you think you may need page and swap files on the system disk, leave the files, but name them PAGEFILE.SPARE and SWAPFILE.SPARE. Don't let the system install them automatically. If you ever need them, you can either install them explicitly, or RENAME them back to .SYS prior to rebooting

[Jansen_8]

Thanks, for the reactions.

I removed the page/swap files on the system disk and the system boots fine. Now I have to let it run for some time to see if the "hang" also keeps away.

Jouk