NetNews Usenet Archive 1992 #20/NN_1992_20.iso/spool/comp/sys/hp/10399

Internet Message Format | 1992-09-14 | 1.6 KB

Xref: sparky comp.sys.hp:10399 comp.unix.admin:5036
Newsgroups: comp.sys.hp,comp.unix.admin,ctp.it
Path: sparky!uunet!mcsun!sunic!psinntp!psinntp!ctp!jmay
From: [e-mail] (Jason May)
Subject: Socket lock-up problem? (HP 9000/847, HPUX8.02)
Sender: [e-mail] ( C News - Network News BBS)
Message-ID: <[e-mail]>
Date: Mon, 14 Sep 1992 13:57:39 GMT
Organization: Cambridge Technology Partners
Lines: 26


We recently were victims of a strange hang situation on our 9000/847.  One
of our users had a process apparently go malloc-berserk, causing heavy
disk swapping and slowing the whole system to a crawl.  However, after
the rogue process was killed, the machine didn't return to normal.
CPU, disk, and memory utilitization (as displayed by HP's Glance tool)
dropped back to low levels, but the load average stayed up at around 9.
Regular UNIX commands (ls, cc, etc.) worked fine, but it became
impossible to launch any X processes; they'd just hang indefinitely.
No error messages were generated.  Established connections to our
Sybase database continued to function fine, but no further connections
were accepted; such connection attempts would also hang.  After rebooting
everything was fine.

Perhaps all socket connection attempts were being denied?  I've never
seen this sort of thing happen before.  If anyone has witnessed this
kind of problem, I'd appreciate any information.

Thanks,
Jason

--
======================================================================
Jason W. May    [e-mail]
CTP, Inc.   304 Vassar St.   Cambridge MA 02139   [phone]
======================================================================