I have a server that no mater what I do , bind will crash every 5 mins , any ideas why or suggestions ?

do you have any log files that we could look at with error messages in? im not 100% familiar with BIND but im sure it will generate error logs in some part the logfile directory on your server.

Douglas you have the same problem we have on one of our servers, and even our top tech's have no clue.

The log's do not report errors, etc it's like a ghost.:cartman:

Edit: This is on a CPanel server correct? if so same situation.. I think it has something to do with CPanel.

What version of bind are you using? There are some recent security issues with versions 4 and 8.

We currently use version 8 across all machines.

Then you should read this:
http://www.cert.org/advisories/CA-2002-31.html

This may or may not be related to your crashing problem, but you should certainly read it.

is it possible you overloaded a box?

this happened to us as there were an excess of processes served

once we dumped a batch of clients to a new box the bind issue vanished.

we came to the conclusions that
1. too many ps on linux makes for weirdness
2. cpanel needs no excuses to act weird, dont give it any more


dos centavos

If you locate the path for the start up script for named [I think it is in /etc/rc.d/init.d, I have forgotten it now] and reload the cache
by

./named reload

it should be fine.

CPanel's built in daemon-monitoring-daemon chksrvrd is not so good at handling named

For a long term solution you can either:
1. Use tinydns and enter your zone records manually OR
2. Use a daemon monitor like monit and set it up to restart bind when ever it dies ; you can slo hack the start up script in monit to include "/path/to/named reload"

I have been using djbdns on FreeBSD for close to a year now and it has not had one single problem [~touching wood~]

Upgrade to BIND 9. 9.0 is the first serious version of BIND written from ground up

Apart from my hosting biz, I also handle support tickets for a very large host with 6000+ domains [most of them CPanel servers] and I can tell you from personal experience that almost 80% of the tickets are on BIND or exim...
... which vindicates my signature anyhoo! ;)

When we had moved our clients to a new server, a parked domainname had replicated itself 3 times in the named.conf, causing the error. So I just deleted it, restarted bind, and then it was fine.

Yes its a cpanel server.

Originally posted by MotleyFool

Upgrade to BIND 9. 9.0 is the first serious version of BIND written from ground up

I'd be careful about saying that, root name servers that run BIND use version 8.3, and they answer millions of queries an hour.

Originally posted by uuallan


I'd be careful about saying that, root name servers that run BIND use version 8.3, and they answer millions of queries an hour.

Yes, *patched* 8.3x name servers, whcih are configured well.

Anyway, I'd suggest upgrading to the newest BIND version. Then show us your logs.

Originally posted by 2host.com

Yes, *patched* 8.3x name servers, whcih are configured well.

Anyway, I'd suggest upgrading to the newest BIND version. Then show us your logs.

Of course, and I would recommend to most people that they upgrade to 9.x as well. However, my point was that the statement that the 9.x version of BIND is not the first serious version of BIND was incorrect.

Originally posted by uuallan


Of course, and I would recommend to most people that they upgrade to 9.x as well. However, my point was that the statement that the 9.x version of BIND is not the first serious version of BIND was incorrect.

Of course. Just noting the vital factor of your comment. Not to insinuate anything to oppose what you said. :-)

Originally posted by uuallan


I'd be careful about saying that, root name servers that run BIND use version 8.3, and they answer millions of queries an hour.

True! :)

[ God I almost sounded like DJB there! ;)]

But it is also true that until 8.x BIND was written by students and that BIND 9 is a rewrite from ground up

http://www.isc.org/ml-archives/bind-users/2000/08/msg01103.html

I have read a few people defending the frequent holes in BIND with a statement similar to what I said , though I couldn't googlemine it in the last hour.

I dont like Oscar Wilde at all.. but I can enjoy his writings. Same case with DJB..

I would love to use his software that is so well designed and architected [and must be written.. I cant review C code for the life of mine!], but if I were to have him as my companion in a long walk I dont think I will enjoy it [atleast of what I see of him from the ISC mailing lists :D]

My humble and foolish opinion is that, in a virtual hosting context, where you are going to have a few hundred zone records and each domain having 5 or 6 records, djbdns is an extremely good choice over BIND

Cheers
Balaji

Originally posted by MotleyFool


My humble and foolish opinion is that, in a virtual hosting context, where you are going to have a few hundred zone records and each domain having 5 or 6 records, djbdns is an extremely good choice over BIND


It certainly is equally as good as BIND, as to whether or not it is a better choice it really depends on the person :D. Personally, I like BIND, and never really dug (no pun intended) djbdns that much...

Bind failing is nearly always due to configuration problems as someone said earlier.,.. check named.conf for duplicate entries and typos such as (domain..com - notice the 2 periods, it happens sometimes) also make sure named.conf isn't pointing to a .db that does not exist in /var/named/ .. although most times named will not start at all if that is the case.. or you get the ndc reload error when trying to restart via whm... If there are no errors then there isn't really any reason it should fail....