This is a known problem with the inetd on Solaris 2.4. A patch 102922-01
should be out sometime, and in the meantime a workaround is to comment
the discard service out of the inetd.conf file. For my setup I have
reverted to serving the Xkernel machines via NFS rather than xfs.
Thanks to: ramey@jello.csc.ti.com (Joe Ramey)
rhmoyer@mmm.com (Robert H. Moyer)
And anyone else whose reply I haven't received yet.
----- Original Message -----
We recently upgraded to Solaris 2.4 on our main SPARCserver 1000.
One of it's functions is to serve Sun 3's running as X terminals
using Xkernel v2.0. This setup worked fine under Solaris 2.3.
This all worked fine under 2.3. Now after about a day, whenever
an X terminal boots and uses the 'clearsockets' program that finds
a random free port for the font serving daemon via discard, the
daemons which inetd spawns to service the request don't die. Truss
shows them repetitively reading from a file descriptor that
corresponds to the discard connection. This leads to about 12 daemons
chewing up all the CPU time and a context switch rate of ~16000/sec.
Even if I kill all these processes, the next time a discard connection
is made the same thing happens. The only (temporary) solution is to
kill and restart inetd, but the problem recurs after about a day.
I believe the reason the problem didn't show up until now is that
none of the X terminals were rebooted since immediately after the
upgrade.
>From /etc/inetd.conf:
discard stream tcp nowait root internal
Has anyone seen this problem, or any problems of a similar nature
with inetd? Any ideas what's going on?
Patches (most from the HW 11/94 CD):
101714-03 101907-04 101933-01 101977-03 102004-01 102039-01 102119-01
101753-01 101910-05 101945-10 101979-03 102007-01 102042-01 102134-01
101829-01 101911-03 101945-13 101981-01 102011-02 102044-01 102169-01
101878-01 101920-01 101950-01 101983-01 102016-01 102048-01 102196-01
101879-01 101920-02 101959-02 101985-01 102020-02 102062-03 102216-01
101880-03 101921-04 101961-07 101987-02 102024-01 102070-01 102226-01
101902-01 101922-04 101969-04 102001-03 102030-01 102105-01 102286-01
101905-01 101923-03 101973-03 102002-01 102035-01 102108-01
101907-02 101925-01 101975-01 102003-01 102037-01 102112-01
Thanks in advance,
Nick Murray
Computer Officer
Department of Computing Science
University of Aberdeen
Scotland
----- End Included Message -----
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:10:15 CDT