edward@comp.vuw.ac.nz (Ed Wilkinson) (09/18/87)
We run Ultrix 1.2 on a Vax 750 & nroff recently told us it didn't have enough core to run. Does anyone have any ideas what to do about this? We are also getting the occasional process kill due to lack of swap space :-( I presume there are some system tables to be adjusted somewhere in the kernel but don't know where. Can anyone help us? Please reply via email as this is probably (for some) a trivial problem. Thanks in advance. Ed Wilkinson ...!uunet!vuwcomp!edward
finegan@uccba.UUCP (09/24/87)
In article <12981@comp.vuw.ac.nz>, edward@comp.vuw.ac.nz (Ed Wilkinson) writes: > We run Ultrix 1.2 on a Vax 750 & nroff recently told us it didn't have > enough core to run. > > Ed Wilkinson ...!uunet!vuwcomp!edward We had problems running reduce on a 11/750 with Ultrix 1.2 and solved all problems by increasing the swap space : In your sys/conf/YOURSYSTEMNAME file is a line like: ..... root on partition swap on partition size sectors ..... just fill in the blanks (where partition is probably (r)ra0a for root and (r)ra0b for swap). The size is not mentioned in the documentation, but the Ultrix v2.0 manual says ~3 times core as max needed (I have heard 2.5 also). When we guessed at 30Mbytes (3+ times our 8M of main memory), you could log in ten or more reduce users, and it would get as slow as you could stand! Also, make sure you are only swapping in one copy of the program. If several people use nroff, only the first one loaded into memory should be used. You can use chmod (sticky bit ?). To recompile the above mess: config YOURSYSTEMNAME, cd .. (where dir of the same name must exist), make depend, make vmunix, save /vmunix, mv /sys/vmunix /vmunix ... This may be off a little, but helps you pick out the part of the installation documentation to look at. You rename the vmunix, in case your new kernel is fubar (to use the latest history lesson), but I don't know if you could then boot either the old one or the new one. I believe you can (I can on SysV 3B2, and I think I had to when Ultrix 1.2 was being installed ...). Whatever. Good Luck - Mike Finegan ...!{hal|pyramid}!uccba!finegan
zemon@felix.UUCP (Art Zemon) (09/28/87)
Under most Unix variants, including Ultrix through the current 2.0 release, all processes are allocated swap space even though they may never actually swap. That means that the total size of all the processes which are running on your system is limited to the SMALLER of your physical memory and your swap space. To bring this closer to home, we have an 8700 with 32 Mb of physical memory and 64 Mb of swap space. Even though we almost never page and *NEVER* swap, I need this much swap space. It is now 1:30pm and the load average is only 2 but I have 38 Mb of swap space allocated (pstat -s give you this figure). Finally, if you are running out of swap space or memory, I suggest you look for programs which gobble large amounts of memory. Use either "sps -la" or "ps -lagx" and look for big numbers. I had been running out of my 64 Mb of swap space and found the culprit was sysline! The dang program reads the entire process table into its own address space just so it can tell you how many processes you have running!! On small machines this is not an unreasonable thing to do but on our 8700, every one of about 50 invocations of sysline was trying to use 750 Kb. I pulled out the offending code and sysline is now down to a more reasonable 132 Kb. I have heard about some new VM implementations that don't allocate swap space until it is really needed. It's about time! There was also a talk at the Phoenix Usenix (by Gould, I think) about a VM system that let you allocate VM up to the sum of swap space plus physical memory. In summary, to get rid of your "killed on swap error" and your "not enough core" error messages, either add physical memory, add swap space, or reduce the size/quantity of the programs you are running. -- -- Art Zemon FileNet Corporation Costa Mesa, California ...!hplabs!felix!zemon
rbj@icst-cmr.arpa (Root Boy Jim) (10/20/87)
From: Ed Wilkinson <edward@comp.vuw.ac.nz> We run Ultrix 1.2 on a Vax 750 & nroff recently told us it didn't have enough core to run. Does anyone have any ideas what to do about this? We are also getting the occasional process kill due to lack of swap space :-( I presume there are some system tables to be adjusted somewhere in the kernel but don't know where. Can anyone help us? Please reply via email as this is probably (for some) a trivial problem. Thanks in advance. I have seen several responses to this problem, and I hope to offer a solution simpler for you. First off, there is nothing that says you must swap on the B partition, it is merely convention. Secondly, you may swap on more partitions than one. In both instances the key is to specify where to swap in the config file. As far as I know, there is no need to specify the size unless you intend to use less of the partition to swap on; the default is to use the whole thing. This would seem to be useful only in building a generic kernel where the root is specified as hp0* or whatever. However, the manual says only to use the clause `swap generic', so I really don't know what `size' is for. Anyway, tell config to swap on *all* your disks; the size increase is negligible, and /etc/fstab controls which ones are actually used. This gives you the flexibility to add extra swap areas without a sysgen. config vmunix root on hp0 swap on hp0 and hp1 and hp2 and ... or config vmunix root on hp0 swap on hp0g and hp1g and ... Ed Wilkinson ...!uunet!vuwcomp!edward (Root Boy) Jim Cottrell <rbj@icst-cmr.arpa> National Bureau of Standards Flamer's Hotline: (301) 975-5688 Did I SELL OUT yet??
roy@phri.UUCP (Roy Smith) (10/23/87)
In article <9863@brl-adm.ARPA> rbj@icst-cmr.arpa (Root Boy Jim) writes: > Secondly, you may swap on more partitions than one. [...] > Anyway, tell config to swap on *all* your disks; Is there any reason why you can't swap on more than one partition on a single disk? I know it sounds stupid (and probably leads to horrible performance) but it might come in handy. For a while, our single RA-81 vax had a 15 Mbyte swap partition (ra0b) and a spare 7 Mbyte partition (ra0e). We were constantly running out of swap space and I couldn't get psyched to doing a full dump/reformat/restore to change the partitions. It would have been nice to be able to say "swap on ra0b and ra0e". Would that have worked? -- Roy Smith, {allegra,cmcl2,philabs}!phri!roy System Administrator, Public Health Research Institute 455 First Avenue, New York, NY 10016
stpeters@dawn.steinmetz (Dick St.Peters) (10/25/87)
In article <9863@brl-adm.ARPA> rbj@icst-cmr.arpa (Root Boy Jim) writes: >Secondly, you may swap on more partitions than one. In both instances the >key is to specify where to swap in the config file. .... >and /etc/fstab controls which ones are actually used. It should probably also be mentioned that additional swap partitions configured into the kernel and listed in /etc/fstab aren't used until a "swapon -a" command is issued, usually in /etc/rc.local. The original posting was about an Ultrix system, and while the default Ultrix rc.local includes the swapon command, not all systems do. -- Dick St.Peters GE Corporate R&D, Schenectady, NY stpeters@ge-crd.arpa uunet!steinmetz!stpeters
chris@mimsy.UUCP (Chris Torek) (10/26/87)
In article <2974@phri.UUCP> roy@phri.UUCP (Roy Smith) writes: >Is there any reason why you can't swap on more than one partition >on a single disk? No. (That is, it should work.) -- In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7690) Domain: chris@mimsy.umd.edu Path: uunet!mimsy!chris