dcr0@GTE.COM (David Robbins) (02/16/90)
My DECStation 3100 has the annoying habit of occasionally becoming thoroughly hung -- totally wedged such that the only possible recovery is to reboot. I am wondering why, and I thought I might ask the net-folk if anyone has observed similar symptoms and has figured out what's going on. The problem is seemingly random, although it usually occurs when I am doing some heavy compiling. The machine will be humming along, banging away at the disk and generally keeping itself busy, when all of a sudden it becomes eerily quiet. No response to the keyboard or mouse, no disk activity, no response to the Ethernet -- it is as though the CPU has simply halted with no way to bring it back to life. My first inclination is to suspect a hardware problem somewhere on the CPU board. However, I can also imagine that there could be a software problem (I'm running Ultrix Worksystem V2.0 Rev. 7) or there could even be a problem stemming from the way Ultrix is installed on the machine. If any of you has seen similar behavior and has found an explanation, I would be pleased to hear it. Otherwise, I guess I'll start complaining to DEC.
grr@cbmvax.commodore.com (George Robbins) (02/18/90)
In article <8331@bunny.GTE.COM> dcr0@GTE.COM (David Robbins) writes: > > My DECStation 3100 has the annoying habit of occasionally becoming > thoroughly hung... > is to reboot. I am wondering why, and I thought I might ask the net-folk > if anyone has observed similar symptoms and has figured out what's going on. > > The problem is seemingly random, although it usually occurs when I am doing > some heavy compiling. The machine will be humming along, banging away at > the disk and generally keeping itself busy, when all of a sudden it becomes > eerily quiet. No response to the keyboard or mouse, no disk activity, no > response to the Ethernet -- it is as though the CPU has simply halted with > no way to bring it back to life. There have been a number of reports about perverse DS3100 behavior when there is a lot of paging/swapping, process creation and not a lot of memory. It seems possible that some of your compilations could be sliding into this mode since the MIPS compiler can do an impressive amount of dancing around your source in an attempt to generate optimized code. Of course it could always be a hardware problem - the only effective way to test this would be to borrow another unit and do your thing there. How much memory do you have? If you could beg/borrow/buy another module or two and that made the problem go away, it would point at Ultrix problems... -- George Robbins - now working for, uucp: {uunet|pyramid|rutgers}!cbmvax!grr but no way officially representing: domain: grr@cbmvax.commodore.com Commodore, Engineering Department phone: 215-431-9349 (only by moonlite)
rsk@boulder.Colorado.EDU (Rich Kulawiec) (02/18/90)
In article <8331@bunny.GTE.COM> dcr0@GTE.COM (David Robbins) writes: > > My DECStation 3100 has the annoying habit of occasionally becoming > thoroughly hung... Here's a similar problem I've been seeing; this one is semi-repeatable. Hook up any old terminal to the serial port on the DS3100; configure it in, log in on it, and run something that generates a decent amount of output, e.g. "yes abcdefg". Now go back to the DS3100 console, log in, fire up a few windows and start clicking the mouse, bouncing between windows, etc. Very shortly, the machine should hang solidly. I'd be curious to know if anyone else has seen this one. -- Rich Kulawiec, rsk@cs.colostate.edu, handel!rsk
GEOMAGIC@kcgl1.eng.ohio-state.edu (Daniel OConnell) (02/20/90)
>My DECStation 3100 has the annoying habit of occasionally becoming >thoroughly hung -- totally wedged such that the only possible recovery >is to reboot. I am wondering why, and I thought I might ask the net-folk >if anyone has observed similar symptoms and has figured out what's going on. >The problem is seemingly random, although it usually occurs when I am doing >some heavy compiling. The machine will be humming along, banging away at >the disk and generally keeping itself busy, when all of a sudden it becomes >eerily quiet. No response to the keyboard or mouse, no disk activity, no >response to the Ethernet -- it is as though the CPU has simply halted with >no way to bring it back to life. My DECStation 3100 exhibited similar behavior during a Fortran (2.0e) compile of a 2000 line main program with level -O2 optimization. After an hour it finally printed the message: Out of swap space and went into a macro funk. The 3100 has 16 Mb of memory! So I tried the same procedure on our DECSystem 5400 which has 32 Mb of memory. The compiler needed 16 free Mb to compile and it took 10 mins. By comparison it only took 13 seconds to compile the same code at optimization of -O1. Conclusion: If your running Ultrix, suspect the software first. Ultrix sucks, but it sucks fast (but sometimes it just sucks air). Dan O'Connell geomagic@kcgl1.eng.ohio-state.edu geomagic@geo1s.mps.ohio-state.edu
litwack@dccs.upenn.edu (Mark Litwack) (02/20/90)
I have heard that a patch is available from DEC that has a new vm_swap.o and vm_text.o on it that is supposed to fix this problem. I have also heard that some problems still exist (like emacs giving segmentation faults) but at least it doesn't lock up anymore. Talk to DEC. -mark