I have an SL5 system that I use for various network polling and 
configuration management functions. About once or twice a day, a telnet 
command, run from an expect script, to a network switch will fail with:

  RTNETLINK answers: No buffer space available

I'm not finding a lot of information about this error. Can someone tell me 
what's happening here? Is there some parameter I can tweak to tune this 
buffer space?

Additional information, that may or may not be relevant, but hopefully 
includes what is relevant:

HP Proliant DL380 G3 system, 2gb memory, 2 CPUs.

This system has one Intel e100 NIC configured as an 802.1q trunk to 
connect with 6 Vlans, plus one other e100 on another network. I've only 
seen the telnet failure going to switches on that one "other" 
(non-trunked) NIC. The two on-board NICs and 6 other e100s are not used at 
present.

Three times a day, the system uses nmap to ping poll each network and 
harvest MAC information. Once/hour it does the telnet thing to each switch 
to collect information. Once/hour does some snmp data gathering. These 
activities do not overlap.

Every once in a while, the nmap ping poll will report several:
   RTTVAR has grown to over 2.3 seconds, decreasing to 2.0
errors while polling one particular trunked network (not the same network 
where the telnet failures occur). 

I tuned some arp table parameters to accommodate the IP polling:
  net.ipv4.neigh.default.gc_thresh1=512
  net.ipv4.neigh.default.gc_thresh2=1024
  net.ipv4.neigh.default.gc_thresh3=1536
I have another very similar virtual SL5 system with the same parameters 
that does not experience the RTNETLINK errors.

[root@xxx ~]# uname -a
Linux xxx 2.6.18-128.7.1.el5 #1 SMP Mon Aug 24 08:22:26 EDT 2009 i686 i686 
i386 GNU/Linux
[root@xxx ~]#

iptables is not used.

                                                                      - 
Bluejay Adametz
 
Definition: Phosflink v. To flick a bulb on and off when it burns out. (To 
fix it?)