| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 427.1 | | QUARK::LIONEL | The dream is alive | Fri Mar 17 1989 10:02 | 12 |
| A 20,000 block pagefile is FAR too small for a 16MB system. On my 13MB
system I am using 70K blocks right now. I was also successful with two
50K block files on two separate disks, though I think that's a bit much.
I suggest boosting your pagefile size to 50K blocks at least, and get rid
of the swapfile - you're just wasting disk space with it.
Also, I believe that the "page file fragmented" message may come out if one
of the files is full, even if you have another one that is empty.
Steve
|
| 427.2 | Need more PAGEFILE | RAMBLR::MORONEY | Anything's legal as long as you don't get caught | Fri Mar 17 1989 10:15 | 12 |
| You're running AUTOGEN yet you still have a 30,000 block swapfile? Betcha you
have SWAPFILE=0 and probably PAGEFILE=0 in your MODPARAMS.DAT. Take these
lines out for now. I have a 1200 block SWAPFILE.
Basically, your problem is your pagefile is too small. DECWindows really eats
up virtual memory. Free up lots of room on your system disk (AUTOGEN will need
room to create a bigger PAGEFILE), remove any SWAPFILE=xxx or PAGEFILE=xxx in
your MODPARAMS.DAT and run AUTOGEN. Remember to purge old SWAPFILEs/PAGEFILEs
after AUTOGENning/rebooting.
-Mike
|
| 427.3 | | STAR::MFOLEY | Rebel without a Clue | Fri Mar 17 1989 11:05 | 5 |
|
And remember to use the FEEDBACK option with Autogen!
mike
|
| 427.4 | | STAR::KAPLAN | Running w/the blade guards disengaged. | Tue Mar 21 1989 07:21 | 12 |
|
When expanding your PAGEFILE it is often useful to keep it contiguous.
I believe its much more important to have the first have contiguous (or
near contiguous) -- it have to do with the algorithms that generate the
fragmentation message(s) to the console.
If you find your disk is too full, you'll probably have to use backup
to defragment your disk.
-jck
|
| 427.5 | | LESLIE::LESLIE | Andy ��� Leslie | Tue Mar 21 1989 08:38 | 8 |
| re: .4 As Steve said in .1, this problem is certainly due to the small
size pagefile.
Don't confuse Pagefile fragmentation with disk fragmentation - they
differ.
Andy
|
| 427.6 | I'm trying a bigger one | FUTURA::MCNULTY | Earth: The hostess, not the meal. | Tue Mar 21 1989 15:51 | 11 |
| Thanks for the replies.
I was using AUTOGEN with feedback. It was trying to make my pagefile *smaller*!
I did a ruthless housecleaning and cleared up a bunch of space on my user disk.
I did a backup/restore of that disk, and created a contiguous 60k secondary
pagefile. I'll see how it goes.
Lance
|
| 427.7 | Still not fixed | FUTURA::MCNULTY | Earth: The hostess, not the meal. | Wed Mar 22 1989 11:14 | 33 |
| That didn't take long. I just got the "Page file space critical, system trying
to continue." message. All my info is the same as in the basenote except my
pagefile setup is now:
Physical Memory Usage (pages): Total Free In Use Modified
Main Memory (16.00Mb) 32768 4373 28092 303
Slot Usage (slots): Total Free Resident Swapped
Process Entry Slots 41 12 29 0
Balance Set Slots 36 9 27 0
Fixed-Size Pool Areas (packets): Total Free In Use Size
Small Packet (SRP) List 570 71 499 96
I/O Request Packet (IRP) List 273 91 182 176
Large Packet (LRP) List 43 22 21 1648
Dynamic Memory Usage (bytes): Total Free In Use Largest
Nonpaged Dynamic Memory 575488 32992 542496 23856
Paged Dynamic Memory 274432 90128 184304 89232
Paging File Usage (pages): Free Reservable Total
DISK$MICROVMS:[SYS0.SYSEXE]SWAPFILE.SYS 30000 30000 30000
DISK$MICROVMS:[SYS0.SYSEXE]PAGEFILE.SYS 1342 -34 1896
FUTURA$DISK2:[SECONDARY]PAGEFILE1.SYS;1 43937 -2345 59992
Any other suggestions/ideas are most welcome. Should I just make the pagefile
still bigger? (Heck, I didn't need to keep any _files_ on this RD54 anyway ;^)
Lance
|
| 427.8 | Don't need it that big in V5 | STAR::MFOLEY | Rebel without a Clue | Wed Mar 22 1989 11:16 | 8 |
|
Why have a 30k swapfile? I highly doubt you need one that big!
Swap the sizes of your primary page and swap files. and
see how you do..
mike
|
| 427.9 | | STAR::MFOLEY | Rebel without a Clue | Wed Mar 22 1989 11:18 | 6 |
|
Also, are you using AUTOGEN with Feedback???
mike
|
| 427.10 | | QUARK::LIONEL | The dream is alive | Wed Mar 22 1989 12:20 | 6 |
| Get rid of the 1900 block pagefile. That's causing most of your problems.
And I agree - you don't need a 30K block swapfile! You probably don't need
one at all, but 5000 blocks tops!
Steve
|
| 427.11 | Another shot in the dark | AGBEAR::HORNER | A.G.Bear, Low tech teddy bear | Wed Mar 22 1989 12:25 | 17 |
| Please don't yell if I don't have this quite right, but...
We had a similiar problem with a small primary page file and a much
larger secondary page file. It seems that some of the VMS processes
get assigned to whatever page file is available at the time that
they start, and this vastly overcommited the primary file. We were
mounting the secondary pagefile disk and installing the secondary
pagefile in our SYSTARTUP_V5 file which was too late in the startup
process. We moved the mounting of disks and installing of the
secondary pagefile into SYPAGSWPFILES.COM which is executed far
earlier in the startup process. This made a tremendous difference
as it seemed to be early enough that the VMS processes were allowed
to use the larger secondary file. Our problem completely disappeared
after doing this.
Dave
|
| 427.12 | | QUARK::LIONEL | The dream is alive | Wed Mar 22 1989 13:16 | 10 |
| Re: .11
The situation you describe was correct for V4, but not V5. In V5, processes
can move around between pagefiles and even can use more than one. Also,
allocation is now made on the basis of percent full rather than absolute
space. So it is now best to have either one big file or multiple files of
similar sizes. Having vastly dissimilar sizes will cause trouble.
Steve
|