greg@umbc3.UMBC.EDU (Greg Sylvain) (11/14/89)
Hi,
I hope someone else has run into this problem before. It goes something
like this :
After about 2 weeks of uptime, the root server doesn't have enough
swap space to start up another login session (an X session particularly).
I ran monitor, and the swap space was 98-99% utilized. And there wasn't
anything running!!! So after a little while of poking around, I gave up
and rebooted. Then I ran monitor again, and the swap utilization was down to
67%.
The best that I can figure is that for some reason, when some processes
die, init can't regain all/or any of that processes alocated swap. Is there
any sort of a cron job that will go out and look for defunct process and
try to regain its swap allocation ? This has also happened on another 340
cluster. Does anyone have any suggestions ?
We're running X11R3 on hp9000/340 workstations.
Thanks in advance,
gregcollin@hpindda.HP.COM (Collin Park) (11/15/89)
I have seen this before when people leave x-windows on the clients running for a while. apparently the program X tends to "grow" its amount of swapspace. Having the clients shutdown and restart x-windows periodically seems to help. I have no information on the source of this anomalous behavior, nor the state of any efforts to get it "fixed." posting only as a user..., ------------------------------------------------------------------------------ Unless explicitly stated otherwise, opinions expressed above are my own and do not explicitly reflect those of the Hewlett-Packard Company or any other person or entity. collin park Hewlett-Packard Company collin%hpda@hplabs.hp.com 19420 Homestead Road -- MS 43LT Cupertino, CA 95014 USA
stroyan@hpfcdq.HP.COM (Mike Stroyan) (11/15/89)
> The best that I can figure is that for some reason, when some processes > die, init can't regain all/or any of that processes alocated swap. Is there > any sort of a cron job that will go out and look for defunct process and > try to regain its swap allocation ? This has also happened on another 340 > cluster. Does anyone have any suggestions ? Swap space is also used by shared memory segments. They can outlive the processes that create them. Perhaps some program on the system is leaving around big shared memory segments. Use "ipcs -mb" to list shared memory segments and their sizes. You can remove shared memory segments with "ipcrm -m <id>" where id is the ID column entry listed by ipcs. Mike Stroyan, stroyan@hpfcla.hp.com
jack@hpindda.HP.COM (Jack Repenning) (11/22/89)
> I ran monitor, and the swap space was 98-99% utilized. Even if your Diskless clients have local swap disks, they can use the rootserver's swap disk as well (I think). So, it's not necessarily the rootserver that's using the space, it might be some client. Use monitor's "k" screen to tell you which system is using the space, and then look on that system for large programs and shared memory segments. (Monitor's "t" screen will be helpful in looking for large programs.) > The best that I can figure is that for some reason, when > some processes die, init can't regain all/or any of that processes > alocated swap. I've never heard of such a problem. But I have experienced problems that caused programs to grow slowly, without bound. As someone else pointed out, the X display server was afflicted with this problem at one point. > We're running X11R3 on hp9000/340 workstations. What revision of HP-UX are you running? If you're running HP-UX 6.2 (and an MIT R3 server), then maybe I can suggest an explanation: there was a rather obscure bug in libmalloc.a in 6.2, which caused our X11 (R2 at that time) server to grow, and would probably afflict a straight-from-MIT one, of any vintage, as well. The work around we found was to link without libmalloc. Or update to 6.5, where the problem is fixed. ------------------------------------------------------------- Jack Repenning - Information Networks Division, Hewlett Packard Company uucp: ... hplabs!hpda!jack or: ... jack@hpda.hp.com HPDesk: Jack REPENNING /HP6600/UX USMail: 43LN; 19420 Homestead Ave; Cupertino, CA 95014 Phone: (408) 447-3380 HPTelnet: 1-447-3380 ------------------------------------------------------------- Disclaimer: These opinions are not necessarily those of my employer. In fact, my manager doesn't know I'm . . . Oh shoot! Here he comes! Quick! Pop up that window with code in it! Bye!
ckw@hpcupt1.HP.COM (Chick Webb) (11/23/89)
>> I ran monitor, and the swap space was 98-99% utilized. > >Even if your Diskless clients have local swap disks, they can use the >rootserver's swap disk as well (I think). So, it's not necessarily >the rootserver that's using the space, it might be some client. I'm not exactly sure what Jack means here, but a diskless client must be set up to *either* swap to the server or to local disk(s). This is set in the kernel configuration file that is used to create the kernel. In the case of a cnode, the swap site must also be indicated in the clusterconf entry for that cnode. Processes running on a cnode will swap to the root server's swap or local swap, but not both! Now, it is possible to start a process that displays on your workstation, but runs on the root server, and thus utilizes the root server's swap that way. I don't think that's what Jack meant, though. >Jack Repenning - Information Networks Division, Chick Webb "Common sense is not so common." Hewlett-Packard Company -- Voltaire Cupertino, CA UUCP: {ucbvax, etc.}!hpda!ckw ARPA: ckw@hpda.HP.COM