[bind-8.2.1] No buffer space available

Jim Reid jim at mpn.cp.philips.com
Tue Jul 27 17:21:37 UTC 1999


>>>>> ">" == C S Chen <cschen at ns1.NCTU.edu.tw> writes:

    >> Hi, One of our DNS servers recently encountered some strange
    >> problem.  -- I'm not sure since when the problem appears and
    >> why.  -- I'd killed the old named process and re-start a new
    >> one, the same problem shows up.  - I'd practices this many
    >> times; removed and/or merged some zone files.  Although fewer
    >> similar messages show up, it seems that the problem is not
    >> fixed.

    >> Did you know how could we avoid and/or fix it ?

    >> Here is the related info: * OS: HP-UX 9.03 * BIND/named: 8.2.1
    >> * system mem: 64M.
    >> Jul 09:42:02 ns named[25282]: ns_req: sendto([168.95.1.2].53): No buffer space available 

Well the ENOBUFS error - "No buffer space available" - occurs when the
operating system has no internal buffers available or some queue of
packets for processing by the network code or device driver is full.
This is usually the sign of a very sick kernel. If the OS can't get
network buffers when it needs them, Very Bad Things happen. Like not
being able to send or receive packets.....

The system might not have been configured with enough memory or buffer
space. It's also possible that there's some buggy code in your kernel
that causes a memory leak so that buffers are not released when they
are no longer needed. You need to talk to the vendor about
that. They'll probably tell to you apply a bunch of patches or upgrade
to an up to date version of HPUX. I believe 9.03 is fairly old. It's
also possible that your system is grossly overloaded with network
activity and is running out of network buffers, but this should be
unlikely. And if it is true, such an overloaded system is not an ideal
platform for DNS.

FWIW the man page for the sendto() system call on my system -
admittedly it's not HPUX - says that an ENOBUFS error is returned
when:
	"The output queue for a network interface was full. This
	generally indicates that the interface has stopped sending,
	but may be caused by transient congestion."

    >> PS. We have another Sun Ultra-I , running Solaris 2.5.1 + BIND
    >> 8.2.1 , with 64M RAM, too.  Up to now, no such problem has been
    >> found.

That tends to point suspicion at the hardware and OS on that HP box.


More information about the bind-users mailing list