cathy@larry.sal.wisc.edu (Cathy Accettura) (07/12/89)
I am having problems tracking down a problem. That we appear to have every afternoon. When I try to do a ps I get not enough memory. When I do a df I get not enough core. I don't see anything in crontab that is causing the problem. I can do a pstat but I don't know what I am looking for. I need some tools to find out what the problem is. I have a MicrovaxII runing Ultrix 3.0. Cathy Accettura University of Wisconsin-Madison Space Astronomy Lab
grr@cbmvax.UUCP (George Robbins) (07/12/89)
In article <209@larry.sal.wisc.edu> cathy@larry.sal.wisc.edu (Cathy Accettura) writes: > I am having problems tracking down a problem. That > we appear to have every afternoon. When I try to do > a ps I get not enough memory. When I do a df I get > not enough core. I don't see anything in crontab that > is causing the problem. I can do a pstat but I don't > know what I am looking for. Generic problems with ps and it's memory peeking friends are usually caused by booting from a kernel other than 'vmunix', such as genvmunix or a new kenel you've generated and not renamed. Your description would seem to indicate other problems, however, probably inadequate swap space. 'pstat -s' is the tool for examinging this. You want to look at the ratio of 'used' to 'free' - the other numbers range from confusing to bogus (it's better in 3.0 than before). You can be eating up swap space if you have users running memory hungry applications like gnu-emacs or lisp, or if you have lots of not-so-big stuff running. Having logins enabled on 32 serial lines, 32 pseudo-ttys and 32 lat pseudo terminals can chew up a surprizing amout of swap space. You can look at the SZ field in a 'ps -aux' output to try to spot who the big hawgs are. If you need more swap space, you *need* more, otherwise all kinds of 'invisible' tasks like sendmail or uucp will be failing due to out of memory problems and your life will become painful and confusing. To add more swap space, you must first have somewhere to put it. Hopefully, you have several disks and there are unused 'b' partitions just waiting for such use. You then need to config a kernel that lists these partitions in "swap on" clauses in the config statment in the config file and make entries for all *except* the first on in fstab, so that the "swapon -a" command in your /etc/rc file will activate the additiona area(s). -- George Robbins - now working for, uucp: {uunet|pyramid|rutgers}!cbmvax!grr but no way officially representing arpa: cbmvax!grr@uunet.uu.net Commodore, Engineering Department fone: 215-431-9255 (only by moonlite)
alan@shodha.dec.com ( Alan's Home for Wayward Notes File.) (07/12/89)
In article <209@larry.sal.wisc.edu>, cathy@larry.sal.wisc.edu (Cathy Accettura) writes: > I am having problems tracking down a problem. That > we appear to have every afternoon. When I try to do > a ps I get not enough memory. When I do a df I get > not enough core. I don't see anything in crontab that > is causing the problem. I can do a pstat but I don't > know what I am looking for. > > I need some tools to find out what the problem is. You're probably running out of swap space. If pstat will run use it look at how the swap space is being used. The command is pstat -s: % psat -s 33440k swap configured 14056k used (3016k text, 0k smem) 19382k free, 4004k wasted, 2k missing avail: 33*512k 6*256k 2*128k 5*64k 2*32k 5*16k 230*1k The first line is the amount of space configured. The 2nd line is how much is being used (and how much of that is text segments and shared memory). The 3rd line is the amount that is available, wasted because of the swap space allocation algorithm and "missing". "Missing" is whatever is left after configured - (used + free). Missing should be small. The last line is how the available space is divided up. It is very easy for a large program to waste a lot of swap space. You probably need to configure more. > > Cathy Accettura > University of Wisconsin-Madison > Space Astronomy Lab Alan Rollow Digitial Equipment Corp. Colorado Springs CO.
collins@unmvax.unm.edu (Bill Collins) (07/13/89)
In article <209@larry.sal.wisc.edu> cathy@larry.sal.wisc.edu (Cathy Accettura) writes: > I am having problems tracking down a problem. That > we appear to have every afternoon. When I try to do > ... [can't allocate memory for commands] If you are running DECwidnows perhaps you should look for masive leaks in the X server and clients. As you use DECwindows, memory leaks and they just soak up memory. If it is, DEC should have taken care of most if not all of the memory leaks in the comming release. > a ps I get not enough memory. When I do a df I get Ps allocates a bit of memory to show processes. Its one of the first to fail. > not enough core. I don't see anything in crontab that Yes, df does some allocation too. Its not cron. > is causing the problem. I can do a pstat but I don't > know what I am looking for. Look for low virtual memory, processes crashing due to memory shortages, etc. Most people who run it around here just reboot. Some go to X11 from MIT. Some cut usage down on DECwindow displays. Some just suffer and will wait until the next version gets here. Have fun, Bill -- collins@unmvax.unm.edu gatech!unmvax!collins
sps@magoo.mcnc.org (Stephen P. Schaefer) (07/13/89)
Just today we saw this problem (ps & df complaining about memory) on a few Vaxstation 2000's with about 16 megabytes of virtual memory running Ultrix 3.0, no DecWindows. You are probably running out of virtual memory. For our variety of processes, the biggest culprit was an X server with SZ > 6 meg of virtual space, according to ps -aux (when we could run ps). I don't understand the server's memory use. We have never observed the server grow smaller, regardless of numbers or size of clients. Workstations with similar (of course, not identical) job mixes, numbers of clients, window managers and up time will have X servers using only 2 meg. I killed the X server and it restarted using a mere 1 1/2 meg; I am now checking it every time something infrequent happens (window glitches and assorted whimsy); it's grown less than 100k in the last four hours. -- Stephen P. Schaefer, Postmaster MCNC sps@mcnc.org P.O. Box 12889 ...!mcnc!sps RTP, NC 27709
graham@fuel.dec.com (kris graham) (07/13/89)
> > ... When I try to doa ps I get not enough memory..... Could be a swap space problem. Up your swap size and tell us what happens. Christopher Graham Digital Equipment Corp Ultrix Resource Center 2 Penn Plaza New York City