wtm@neoucom.UUCP (11/09/87)
#ifdef lineater A while back somebody posted a suggestion of adding more gettys in /etc/inittab to permit several logins at the same time on the console. I tried the suggestion, and it seemed to work for a while. After a few days, the machine crashed inexplicably. The screen still displayed, the mouse pointer moved, and the modem would go off hook in response to an incomming call. The machine would simply ingore anything typed in over the console or modem. On the command line at the bottom of the screen there was a message to the effect, "there isn't enough memory. close some windows." Hmmmm.. Curious, there is only one window open. Several days later, suddenly, I got about the same effect. The console started going at about the rate of a 300 baud terminal. Next came the "close some windows" message. Click on the X in the corner has no effect. (a total of 3 windows were open) Sigh, and reboot. I examined /usr/admin/unix to look for some hideous error, but nothing beyond the floppy disk error from the unreadable diskette I got in the document preparation set. Next, try an fsck-- nothing out of the ordinary reported. I came to the conclusion it must have something to do with the changes to /etc/inittab. Well, I returned /etc/inittab back to the default configuration, and everything has been fine since. By the way, the above effects were experienced running v 3.51. Has anybody else experienced anything akin to this? I'm still worried that the MMU might be out to lunch, not deallocating memory when a task finishes. Also, does anybody know the significace of the green boxes that print in the upper left hand corner when power is first applied? I notice that when booting from a complete power-down, six boxes are printed. A reboot from running shutdown only prints one or two boxes. As usual, thanks for the help I've received from the network. Bill Mayhew Division of Basic Medical Science Northeastern Ohio Universities' College of Medicine Rootstown, OH 44272-9989 USA phone: 216-325-2511 (wtm@neoucom.UUCP)
rmr@ncrpcd.UUCP (Roscoe) (11/10/87)
I have been running three getty's on my console for several months now, and have not had the problems you meantioned. ( unix 3.5) Although, it does create some annoying problems with the windowing done by the telephone manager. The telephone manager does not consider itself active, unless someone is logged into the original console window. (or at least thats what i believe it wants). I have a question/problem. Since I am running several getty's from my console. And the telephone manager uses the original console as its base. Is there anyway to convince the telephone manager to always use the same "phone book/directory" and the same history file without having to log the same account into the original console window. Another question I have is, is there anyway to get rid or hide all the Notes/Filcabinet/Wastebasket,etc directories that are put all over the place. Is this from the wmgr, or the smgr? Making them "." directories would be acceptable. Somewhere out of mind, Roscoe Rush ncrpcd!rmr
brant@manta.UUCP (Brant Cheikes) (11/11/87)
In article <750@neoucom.UUCP> wtm@neoucom.UUCP (Bill Mayhew) writes: >A while back somebody posted a suggestion of adding more gettys in >/etc/inittab to permit several logins at the same time on the >console. [...] I tried the suggestion, and it seemed to work for a >while. After a few days, the machine crashed inexplicably. [...] >I came to the conclusion it must have something to do with the >changes to /etc/inittab. [...] I returned /etc/inittab back >to the default configuration, and everything has been fine since. It's an interesting theory, but it certainly isn't reproduced on my machine. I'm also running 3.51 and have added an extra console to /etc/inittab. This configuration has been working for several months. How many extra consoles did you add? Maybe there's some upper bound. I tend to doubt it's a hardware problem, since I can't see how restoring inittab would have relieved such a failure. Nevertheless, I suppose you've run the diagnostics (for whatever they're worth)? -- Brant Cheikes University of Pennsylvania Department of Computer and Information Science ARPA: brant@linc.cis.upenn.edu, UUCP: ...drexel!manta!brant
jeff@cjsa.UUCP (C. Jeffery Small) (11/12/87)
In article <750@neoucom.UUCP>, wtm@neoucom.UUCP (Bill Mayhew) writes: > #ifdef lineater > > After a > few days, the machine crashed inexplicably. The screen still > displayed, the mouse pointer moved, and the modem would go off hook > in response to an incomming call. The machine would simply ingore > anything typed in over the console or modem. On the command line > at the bottom of the screen there was a message to the effect, > "there isn't enough memory. close some windows." Hmmmm.. > Curious, there is only one window open. > I too have experienced a similar problem on a new 3B1 running 3.51. Typically, I will log into my Office and enter UNIX. As the ksh starts up, my system often hangs completely. Note: (Sometimes I make it into the k-shell and execute a few commands before the system hangs. Other times the system hangs somewhere between login: and Enter Password: !!) As Bill reported above, the only responses you can get from the kernel are: - Mouse pointer movement - Moving/resizing Office window You cannot do the following: - Get a shell prompt - Close the Office window - Use Help - Access the Window [W] Icon - Access the phone manager ... etc - Cannot login on remote terminal - no getty(s) running. However, if you click the mouse at various points on the Status line, you can get the following error message (like Bill reported) to appear: Error - can't create window - Touch ENTER to continue. The only recourse is to reboot the system. There is no indication in the adm logs of a problem and I have been unable to locate any other signs of distress. After reboot, the system runs fine for one or more days and then the problem repeats. I have three 3B1 systems and this problem only occurs on one machine. The suspect machine has a 2-port expansion board and is used as the primary machine for uucp (usenet) feed. These are the only real differences of note. Unlike Bill, I have made no modifications to inittab and I do not have any special programs running which might burn up system resources. Anyone out there have any insight into this one! Any suggestions will be appreciated. -- Jeffery Small (203) 776-2000 UUCP: uunet!---\ C. Jeffery Small and Associates ihnp4!--- hsi!cjsa!jeff 123 York Street, New Haven, CT 06511 hao!noao!---/
andys@shlepper.ATT.COM (a.b.sherman) (11/15/87)
In article <47@cjsa.UUCP>, jeff@cjsa.UUCP (C. Jeffery Small) writes: > [Deleted description of a very sick 3B1] > the problem repeats. I have three 3B1 systems and this problem only occurs > on one machine. The suspect machine has a 2-port expansion board and is > used as the primary machine for uucp (usenet) feed. These are the only I have never seen the problem, but I have read on the net that there is one version of the combo board that crashes the 3B1 when a break is received. Could that be your problem. Other folks who have saved past postings could contribute the revision level that's bad. -- Andy Sherman / AT&T Bell Laboratories (Medical Diagnostic Systems) 480 Red Hill Road / Middletown NJ 07748 / (201) 615-5708 UUCP: {ihnp4,allegra,akgua,cbosgd,mtune....}!shlepper!andys INTERNET: andys@shlepper.ATT.COM
alex@umbc3.UMD.EDU (Alex S. Crain) (11/15/87)
In article <750@neoucom.UUCP> wtm@neoucom.UUCP (Bill Mayhew) writes: >A while back somebody posted a suggestion of adding more gettys in >/etc/inittab to permit several logins at the same time on the >console. [...] I tried the suggestion, and it seemed to work for a >while.... I've been running 3 gettys for a while now and have had no crashes, although the machine does seem to be a trifle confused at times. The most obvious symptom is the message "You cannot use the Telephone manager without logging in.", which appears occationaly for no apperent reason. I also have interminate problems with execl() failing with the error Not a tty, but this may be an application problem. :alex.
jr@amanue.UUCP (Jim Rosenberg) (11/16/87)
In article <750@neoucom.UUCP> wtm@neoucom.UUCP (Bill Mayhew) writes: >#ifdef lineater > >A while back somebody posted a suggestion of adding more gettys in >/etc/inittab to permit several logins at the same time on the >console. > >I tried the suggestion, and it seemed to work for a while. After a >few days, the machine crashed inexplicably. The screen still >displayed, the mouse pointer moved, and the modem would go off hook >in response to an incomming call. The machine would simply ingore >anything typed in over the console or modem. On the command line >at the bottom of the screen there was a message to the effect, >"there isn't enough memory. close some windows." Hmmmm.. >Curious, there is only one window open. > > ... other problems ... I think I'm the one that posted the idea of multiple gettys. You will consume one window for each getty, *regardless* of whether anybody is logged in on that getty. If one or more of those logins is running ua, it *is* quite possible to run out of windows. I typically run 6 gettys on /dev/window, have been doing this for months, and have had no problems. I did suffer one mystery crash, but I can't attribute it to any window stuff. However, to run most ua things I would have to blow off at least two of the gettys; it seems to me that even 4 once ran me out of windows. (Flame on:) I must say that I find ua absolutely **PROFLIGATE** in its use of windows. There are programs for Mess-DOS that have heaps and gobs of windows with no multitasking at all. Why should ua consume a scarce resource, i.e. one of only 12 /dev/w* devices, just to tell you, OK, bunky, you can press RETURN now and the world won't end? What ua needs is a library to allow subwindows within a given /dev/w* device so that fewer of them are used. Consuming a new /dev/w* should be like forking a new process: you do it when you have to, but you'd prefer to get as much done as possible with one process. A message sent by a program is logically part of that program; you should only spawn a new window when the prospect of *CLICKING BACK AND FORTH* between parent and child windows is a reasonable way for the program to be operated. An extra /dev/w* in place of the typical Press Return to Continue ??? Sheesh! -- Jim Rosenberg CIS: 71515,124 decvax!idis! \ WELL: jer allegra! ---- pitt!amanue!jr BIX: jrosenberg uunet!cmcl2!cadre! /
wtm@neoucom.UUCP (Bill Mayhew) (11/16/87)
<<Discussion about multiple console window gettys not working>> I've had mixed results on this one. I got sevral letters to the effect, "I've been doing this for the last year, and never had a problem". I also received several letters to the effect, "I had the same results as you." Perhaps, this is a hardware difference? I forget what version of hardware gets displayed at the boot-up at the moment. I was playing it very cool and only had one extra console window. I got the error message when only one window was awake. It seems that when the getty respaws, it grabs new memory for the window, but doesn't free the previous memeory. What is odd is that this does not seem to happen with only one console window. Perhaps, this could be a problem in a shared library used for window management. It could be that oddness happens if the idle getty's memory gets swapped out. I'm not sure if this really happens; I've had too many other irons in the fire to get a chance to check this out. Diagnostics doesn't report anything out of the ordinary. -Bill
jeff@cjsa.UUCP (C. Jeffery Small) (11/16/87)
In article <169@shlepper.ATT.COM>, andys@shlepper.ATT.COM (a.b.sherman) writes: > In article <47@cjsa.UUCP>, jeff@cjsa.UUCP (C. Jeffery Small) writes: > > [Deleted description of a very sick 3B1] > > the problem repeats. I have three 3B1 systems and this problem only occurs > > on one machine. The suspect machine has a 2-port expansion board and is > > used as the primary machine for uucp (usenet) feed. These are the only > [ Andy Sherman / AT&T replies ...] > I have never seen the problem, but I have read on the net that there > is one version of the combo board that crashes the 3B1 when a break > is received. Could that be your problem. Other folks who have > saved past postings could contribute the revision level that's bad. Good thought. However, I checked with the hotline about this problem and received a replacement IC for the board. After the upgrade the problem is still with me. Today for example, I arrived to find my status line message area filled with a solid reverse video bar!. I logged into the Office and entered Unix. Got a ksh prompt and typed "ps -e" which did not echo!!! - but the command ran. Typed "ps -e" again - no echo but command executed. Typed "ps -e" a third time and system hung and I had to reboot :-(. I never did get to try and see what Msg might have told me. After reboot, everything works fine once again with no signs of the problem anywhere. Anyone have any other clues? -- Jeffery Small (203) 776-2000 UUCP: uunet!---\ C. Jeffery Small and Associates ihnp4!--- hsi!cjsa!jeff 123 York Street, New Haven, CT 06511 hao!noao!---/
ford@crash.cts.com (Michael Ditto) (11/21/87)
I have had two gettys running on /dev/window for a few months. I have encountered only one problem: getlogin(3C) does not work correctly for the second getty (whichever one is last in /etc/inittab). So, for example, when running hack, my name is shown as "LOGIN" if I happen to be on the "second" getty and the first one is idle. Since there is no easy way to distinguish the two, it is difficult to log in on the "correct" window. It doesn't bother me enough to outweigh the value of that "emergency" getty in the background, so I have left it. I don't log in on the console very much, though. -- Mike Ditto -=] Ford [=- P.O. Box 1721 ford%kenobi@crash.CTS.COM Bonita, CA 92002 ford@crash.CTS.COM
lynn@engr.uky.edu (H. Lynn Tilley) (11/24/87)
>The only recourse is to reboot the system. There is no indication in the >adm logs of a problem and I have been unable to locate any other signs of >distress. After reboot, the system runs fine for one or more days and then >the problem repeats. The error message you are getting and the fact that rebooting clears the problem up may indicate that the problem stems from having excess tmp files or very large log entries. The 7300's (3b1's) keep a complete log of all uucp traffic and this file can grow very large, very quickly. I believe that the file is cleaned once a week by the machine and also when the machine is rebooted. Good luck. -- | Henry L. Tilley UUCP: cbosgd!ukma!ukecc!lynn | University of Kentucky CSNET: lynn@engr.uky.edu V Engineering Computer Center BITNET: lynn%ukecc.uucp@ukma O voice: (606) 257-1752 ARPANET: lynn@a.ecc.engr.uky.edu
allbery@ncoast.UUCP (11/29/87)
As quoted from <2031@crash.cts.com> by ford@crash.cts.com (Michael Ditto): +--------------- | I have had two gettys running on /dev/window for a few months. I have | encountered only one problem: getlogin(3C) does not work correctly | for the second getty (whichever one is last in /etc/inittab). So, | for example, when running hack, my name is shown as "LOGIN" if I happen | to be on the "second" getty and the first one is idle. Since there is | no easy way to distinguish the two, it is difficult to log in on the | "correct" window. +--------------- The fix is quite easy: link /dev/window to another name or two, such as /dev/window1 and /dev/window2; then specify a different name for each of the getty's. Then getlogin() can easily determine the correct window. -- Brandon S. Allbery necntc!ncoast!allbery@harvard.harvard.edu {hoptoad,harvard!necntc,cbosgd,sun!mandrill!hal,uunet!hnsurg3}!ncoast!allbery Moderator of comp.sources.misc
allbery@ncoast.UUCP (Brandon Allbery) (12/02/87)
As quoted from <1810@ukecc.engr.uky.edu> by lynn@engr.uky.edu (H. Lynn Tilley): +--------------- | >The only recourse is to reboot the system. There is no indication in the | >adm logs of a problem and I have been unable to locate any other signs of | >distress. After reboot, the system runs fine for one or more days and then | >the problem repeats. | | The error message you are getting and the fact that rebooting clears | the problem up may indicate that the problem stems from having excess | tmp files or very large log entries. The 7300's (3b1's) keep a complete +--------------- I was using my 3B1 yesterday to do word processing, so I turned on the printer and then left it on later. A few hours after I was done printing, the system suddenly slowed down and smgr began reading the disk and updating the status line once per second. Also, the "windaemon" process (pid == 3) began racking up enormous amounts of CPU time. This went on for about an hour, during which I checked everything I could think of; then gave up and went to bed, expecting the computer to have crashed in the morning... got up again and shut off the printer. And the computer quieted down. I let it sit for a few minutes, then powered up the printer again... the disk started "ticking" again, once per second, and smgr began updating once/second again; and it stopped when I powered the printer down again. For those interested, I use a Tandy DMP105 (for final copies, I create a floppy and print it on my boss's LaserJet+). Anyway, I suspect that the parallel printer driver sends some unexpected interrupts which tickle the kernel in some strange ways. It is possible, given the effect on "windaemon"'s CPU time, that this might be a cause for peoples' spurious "too many windows open" crashes as well. -- Brandon S. Allbery necntc!ncoast!allbery@harvard.harvard.edu {hoptoad,harvard!necntc,cbosgd,sun!mandrill!hal,uunet!hnsurg3}!ncoast!allbery Moderator of comp.sources.misc