Date: Sat, 23 May 1998 01:00:25 -0400 (EDT) From: CyberPeasant <djv@bedford.net> To: r-beer@onu.edu (Robert Beer) Cc: freebsd-questions@FreeBSD.ORG Subject: Re: Freeze Up on 2.2.2 Message-ID: <199805230500.BAA23111@lucy.bedford.net> In-Reply-To: <l03102814b18b5319ca5a@[140.228.15.35]> from Robert Beer at "May 22, 98 12:26:17 pm"
next in thread | previous in thread | raw e-mail | index | archive | help
Robert Beer wrote:
> I have a 2.2.2 host that has been successfully serving about 40000 e-mail
> messages daily for about five months without a reboot. This has happened a
I hope any NT admins reading this get the picture: Mr. Beer gets a glitch
twice a year and is /worried/.
> few times before but the time has come to delve deeper into this problem.
> Only adminstrators log into this host.
>
> No information was logged in the various log files or on the console. The
> only strange item was that logging to syslog files in the /var partion
> stopped about 2 hours before the computer became unresponsive. The mail
> spool and mail queue are on different disks. Therefore I am pretty sure
> disk space was not the problem as /var has about 80 MB free.
Agree. You would have probably seen something in the logs. I can't imagine
a fs-full problem that would hang a root console login.
> I became aware of the problem when I grep'ed a popper log file for a
> username and got no response or prompt back. The host was still pingable.
> The alternate windows of the console worked and I tried to log in on the
> second one. The username echoed but the password prompt never appeared
> after the username was entered. There was some system disk activity but
> all disk activity soon stopped. After about 10 minutes of looking around I
> flipped the power switch.
Yeah. I'll bet some other network services were still working, too.
Like daytime. Keep watching ... you'll see occasional disc activity,
maybe. (disks synching, etc). The system looks alive, but won't talk to
you. I have a feeling that it is unable to fork new processes: old
processes still work. (that getty, for example, obligingly offering
a login: prompt, but, I theorize, unable to fork "login" to get the
password.) This suggests the the system is out of process slots or
some other kernel resource. There are DoS attacks that will cause
some heavy forking, I think.
> This is a Tyan D1662 with a 200 Mhz PPro and 96 MB of EDO memory and two
> NCR 875 disk controllers. I have a couple other systems that are VERY
> similar.
>
> Does anybody have any suggestions?
This sounds familiar. I've seen systems in this very state. It
sounds like what happens when NFS gets hosed. I think I have seen
this on Host A, when Host A has mounted a NFS partition (writable,
IIRC) from Host B, then Host B crashes for some unrelated reason
(cosmic rays), and Host A tries to umount that partition.
Next time it happens, try running a packet sniffer (tcpdump) on another
host on the line, and see if the packets in/out of the hung host have
any tell-tales.
I think there are some DoS attacks that will show this kind of symptom,
too.
Dave
--
Unix System 7:
an improvement on all other Unix releases, previous and subsequent.
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199805230500.BAA23111>