BIND 9.2.0----exiting due to assertion failure

[Rgomes]

Hi All,

I got this meesage in syslog:

# tail -f /var/adm/syslog/syslog.log
Jan 26 14:25:22 dns2 named[3789]: client 64.29.144.44#33311: query:
dns2.bttb.ne
t IN AAAA
Jan 26 14:25:22 dns2 named[3789]: client 207.126.105.146#1040: query:
dns.bttb.n
et IN AAAA
Jan 26 14:25:23 dns2 named[3789]: client 207.126.105.146#1040:
query 'dns.bttb.n
et/IN' denied
Jan 26 14:25:22 dns2 named[3789]: client 64.29.144.67#35309: query:
dns2.bttb.ne
t IN AAAA
Jan 26 14:25:23 dns2 named[3789]: client 64.29.144.44#33311:
query 'dns2.bttb.ne
t/IN' denied
Jan 26 14:25:23 dns2 named[3789]: /ux/core/inet/src/bind.src/bind-9/bind-
9.2.0/l
ib/isc/task.c:395: REQUIRE((((task) != 0L) && (((const isc__magic_t *)
(task))->m
agic == ((('T') << 24 | ('A') << 16 | ('S') << 8 | ('K')))))) failed
Jan 26 14:25:22 dns2 named[3789]: client 64.29.144.67#35309:
query 'dns2.bttb.ne
t/IN' denied
Jan 26 14:25:23 dns2 above message repeats 2 times
Jan 26 14:25:23 dns2 named[3789]: exiting (due to assertion failure)
Jan 26 16:36:02 dns2 sshd[7285]: Accepted password for root from
209.58.24.239 p
ort 3752 ssh2
# ps -ef|grep named
root 7308 7288 0 16:36:57 pts/0 0:00 grep named

Jan 26 14:25:23 dns2 named[3789]: exiting (due to assertion failure)===>
What is it??

Can anybody help?

Thanks in advance,
Richard

[Steven E. Protter]

Have you got the latest security patches for BIND from http://software.hp.com
?

Is the request being made a valid one?

Have you checked the accuracy and integrity of the zone records in your DNS database?

Any of these issues can cause this problem.

More details about your environment are needed to help. First instinct though always is are you patched up completely.

There are DNS related patches in the patch database link off the main page of http://itrc.hp.com

SEP

[Rgomes]

Hi SEP,

BIND ver is 9.2.0, it had been running quite fine for last 3 months. It is our primary dns server but it is not working now. We have one backup dns server and it is online now.

I am checking latest dns patches, if I find some, will install.

Interesting observation from nslookup( attachment)

" QUESTIONS:
15.194.112.203.in-addr.arpa, type = PTR, class = IN

------------
recvfrom: Protocol error
send: Protocol error
recvfrom: Protocol error
send: Protocol error
SendRequest failed
*** Can't find server name for address 203.112.194.15: No response from server"

I guess, reverse-file is not OK?? I don't know exactly....need to dig more.

Thanks for your reply,
Best regards
Richard

[Todd Whitcher]

Hi Richard,

I did a quick search of the lab database, I do see some named issues w/ assertion failues but they are for different parts of the code, resolver.c and named.c but not task.c. Someting in the task.c is whats causing your issue here.

/ux/core/inet/src/bind.src/bind-9/bind-
9.2.0/lib/isc/task.c:395: REQUIRE((((task) != 0L) && (((const isc__magic_t *)
(task))->m

The problems occured in testing due to either stress testing or mis-configuration or mis-use of of rndc commands.

There appears to not be a bug fix for this, what I suggest is you open a case with the Response Center and provide the following.

#what /usr/sbin/named
#named.conf
#syslog.log named messages
# swlist -l product output.

Todd

[Rgomes]

Hi Todd,

Thanks a lot for your advise. I already logged a case with resp. center and still working. The problem still persists and getting more and more complex( I guess).

I did not do anything with rndc, but don't know about any mis-config of any zone file.

Thanks again and best regards,
Richard

Oh, his name is SENIN,SHERMAN (Philippines,ex1). You may want to chat with him.... :)

[Geoff Wild]

I think this is a "thread" issue - happened in Bind 9.1.3?

Did you recently upgrade?

Try latest patches - or if you dare, download/compile Bind from source....current version is 9.2.3

Not too sure when HP will have that....

Could be an SMP issue - try running it with the extra option "-n 1". This option will have BIND run single-threaded, and only
use one cpu.

This might help:

http://www1.itrc.hp.com/service/cki/docDisplay.do?docId=HPSBUX0207-202

Install version BIND920v2.depot

Rgds...Geoff

[Todd Whitcher]

Richard,

I sent Senin a message regarding the issue. I gave him some contact information and let him know he can contact me for assistance. Chances are the lab will have to take a look.

Todd

[Rgomes]

Hi Todd,

Thats great indeed!! It's great to hear that you talked to Senin about this problem. Wao that's wonderful.

One thing I like to ask to BIND expert like you guys:

Is there any bug related to sniffer( type of programm) which uses one TCP port and for which BIND 4.x.x dies. This bug is inherited to BIND 9.2.x??

Thanks to you all,
Richard

[Rgomes]

Hi Geoff,

My named service is running in single thread. I don't know but there were some changes( actually adding one www entry to respective zone file) in one of my zone file, after that I am having this problem.

I will update the BIND ver to latest, but first want to hear from HP lab, if they have anything to recommand.

Best regards, and thanks
Richard

[Todd Whitcher]

Hi Richard,

The latest is Revision 5, check your what output.

what /usr/sbin/named
/usr/sbin/named:
$Revision: 5.0 $ Tue Dec 2 08:08:55 GMT 2003
named 9.2.0 Tue Dec 2 08:08:55 GMT 2003
Copyright (C) 1995-1998 Eric Young.All rights reserved.

If you dont have Rev. 5.0 load that from software.hp.com.

I dont see any issues w/ running a sniffer on the network. I'd be suspicious if your network admins were running Security Port SCans and every time they ran it you had this problem :-).

Todd

[Rgomes]

Hi Todd,

Thanks for your mail and feedback.

# what /usr/sbin/named
/usr/sbin/named:
$Revision: 2.0 $ Sat Sep 21 11:37:57 GMT 2002
named 9.2.0 Sat Sep 21 11:37:57 GMT 2002
Copyright (C) 1995-1998 Eric Young.All rights reserved.
#
SO, I guess I have to update it.
Another issue, pls see the below and give me light on it( some plicy related matter??)

# nslookup
Using /etc/hosts on: dns2

> server 203.112.194.15
Specifying a server has overridden the switch policy order.
The reset command will reinstate the order specified by the switch policy.
Default Name Server: dns2.bttb.net.bd
Address: 203.112.194.15

> yahoo.com
Name Server: dns2.bttb.net.bd
Address: 203.112.194.15

Trying DNS
Non-authoritative answer:
Name: yahoo.com
Address: 66.218.71.198

Thanks,
Richard

Note: if possible pls give me mail to richardflora19@yahoo.com

[Rgomes]

Hi All,

Problem has been solved. There was a mis-configuration in one of my db file. I recreated it, after that it seems OK. :)

But still want to update to Rev 5.

Thanks all of you for your suggestions and advise.

Richard

[Ramesh Velu]

Hi Richard,

I am facing the similar problem with my DNS server running BIND 9.2.0 and named Revision 4.0.

The Message in syslog.log file is
************
May 3 17:31:55 dns named[4915]: /ux/core/inet/src/bind.src/bind-9/bind-9.2.
0/lib/isc/task.c:395: REQUIRE((((task) != 0L) && (((const isc__magic_t *)(task))
->magic == ((('T') << 24 | ('A') << 16 | ('S') << 8 | ('K')))))) failed
May 3 17:31:55 dns named[4915]: exiting (due to assertion failure)
*************

Any suggesstion to fix this problem?

Cheers, Ramesh

[Todd Whitcher · Accepted solution]

Hello Ramesh,

There have been a few issues with named exiting due to assertion failures. In Richards case it was due to a configuration issue. Check the syslog.log file for named errors on any of your zone data files.

The HP lab for named has also determined that this issue occurs more often on multi-cpu systems.

workaround
If you are experiencing this problem an a server with
multiple cpu's, then it may help to execute named with the -n option
set to 1.

# named -n 1


-n no_of_cpus
This option can be used to create worker threads equal
to no_of_cpus to take advantage of multiple CPUs. If
no option is given, named will try to determine the
number of CPUs present and create one thread per CPU.
If named is unable to determine the number of CPUs, a
single worker thread is created.


If you do not isolate a configuration issue in your zone files I would suggest you contact the HP response Center so that the lab can investigate your issue.

The defect id for that task.c problem is
JAGaf09745.


Hope that helps,

Todd

[Ramesh Velu]

Todd,
Thanks for yout reply, my DNS hosts has 2 cpus, I have restarted named with -n option and I am monitoring the named process.
If named dies agian I'll raise a call with HP support.

Cheers, Ramesh

[Rgomes]

Hi Ramesh,

Did not see your post earlier. I hope Todd has put you in the right track.

regards,
Richard