wierd 'growreg' and 'getcpages' errors ISC 2.2 from tcp?

[Stu Donaldson]

I have been getting some unusual errors in my syslog output.

Apr  5 14:11:40 hindmost unix: tcpcmds: 
	non-fatal error in unknown event: 0000000B

Apr  5 14:11:41 hindmost unix: NOTICE: growreg - 
	Insufficient memory to allocate 52652 pages - system call failed

Apr  6 07:17:35 hindmost unix: NOTICE: getcpages - 
	waiting for 1 contiguous pages

The growreg and getcpages errors ocasionally show up with a non-fatal
error from tcpcmds as in the above case.  They are always related to
receiving an incoming ftpd/telnetd/rlogind connection.  Ocassionaly
the messages are repeated several times.  However, the incoming
connection always seems to work.

Here is the output from netstat -m.  Occasionally there will
be something in the 'fail' column, but the Max is usually well
below the allocated, so I suspect it is trying to allocate
a large number of blocks at once and failing, rather than
trying to allocate them on the way up.

		 alloc	 inuse	   total     max    fail
streams:	   160	    63	   46039      87       0
queues: 	   640	   352	  272872     492       0
mblocks: 	  2327	   435	 9915274     746       0
dblocks: 	  1862	   417	 8609838     743       0
dblock class:
    0 (   4)	   256	     8	  151170     205       0
    1 (  16)	   256	    36	  829327     207       0
    2 (  64)	   384	   155	 5944361     170       0
    3 ( 128)	   600	   194	  562355     271       0
    4 ( 256)	   128	     3	  215176      78       0
    5 ( 512)	   128	    10	  227163      20       0
    6 (1024)	    50	    11	  407849      38       0
    7 (2048)	    30	     0	  270303      27       0
    8 (4096)	    30	     0	    2134      11       0

This system does have NFS installed, and several remote mounts
to it. It also only has 8 Mb of memory, and frequently gets 10
or more users logged in at the same time.

Any suggestions on how to fix this problem, would be appreciated.

	-- Stu --