BIND 9.5.0-P2 crash on solaris 8 : socket.c:2586: fatal error:

[Fr34k]

Hello,

What does "ulimit -a" say for the shell and user that named process is running in?
open files should be >= 8192

I am not sure if this is the culprit, but may be something to check.

Note that SUN has Solaris8 on the chopping block:
http://www.sun.com/service/eosl/solaris/solaris_vintage_eol_5.2005.xml
I mention this as leverage, should it be necessary, to move to something newer.

Otherwise, there are many tunable parameters in Solaris8 that improve TCP/UDP/IP which appear to be built into Solaris9+ (for the most part).
One can do some googling on the many resources for this should that be desired.

Hope this helps.

 


----- Original Message ----
From: Jeffrey Collyer <jwc3f at virginia.edu>
To: Bindlist <bind-users at isc.org>
Sent: Monday, September 8, 2008 12:56:13 PM
Subject: BIND 9.5.0-P2 crash on solaris 8 : socket.c:2586: fatal error:

Running bind 9.5.0-P2 with -DISC_SOCKET_FDSETSIZE=8192
on Solaris 8, recently patched, on Sparc.  Server is DNS only, on 
reasonably old hardware, and thus fairly busy most of the time.


Got this in the the logs :

named[9706]: [ID 873579 local4.crit] socket.c:2586: fatal error:
named[9706]: [ID 873579 local4.crit] select() failed: Invalid argument
named[9706]: [ID 873579 local4.crit] exiting (due to fatal error in library)

and named had to be restarted.  I know that there have been a few others 
reporting this on the list, but I wanted to add one to the frequency 
counter.

We did get a good core file dumped, if that would be of any help in 
debugging this.


Jeff
-- 
Jeffrey Collyer
University of Virginia