[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: DNS cpu utilization

I would suggest you setup some sort of monitoring on the bind process. I personally use mrtg to graph the output of "rndc stats". The main dns server I am responsible for runs about 12k-15k dns queries per 5 minutes and has just about zero load. It's running on a slowish (aren't they all) sparc.

I think your bind process is probably not configured optimally for your machine. You might want to increase the amount of ram available to the bind process (max-cache-size 50M;) and setup some server-side monitoring so you can see what the machine is actually doing.

Is the machine hitting the swap at all? You DON'T want a dns server swapping out to the disk all the time. Is the machine running any other services? You said dns was this machine's primary service, but what else is running? Close down all of the services except ssh and bind and see how it runs.

If you "push a machine too hard" you should not have ANY stability problems. If you have stability problems anytime, you have faulty hardware. If you start running the limit of your hardware, the machine will let you know by dropping dns requests and being overall unresponsive.


peace bwitchu wrote:

These two bind servers are authoritative for 200 zones
and as far as clients go that's a hard one.  Because
of the way everything snowballed before I got ahold of
it it could be as many as > 10,000 this is a best
guess of course.  Once I prove that debian is the way
to go with this I plan to implement the split dns
topology.  I don't have any statistics yet but plan to
have some by week end.

--- Russell Coker <russell@coker.com.au> wrote:
On Thu, 8 May 2003 03:08, peace bwitchu wrote:
this box working too hard or is the normal.  Since
this box is dedicated solely to dns I just want to
make sure that I'm not pushing bind too hard and
up with stability problems.
Without knowing how many machines are using "this
box" as a DNS proxy or how important the domain(s) it serves are how can we determine if it's normal load?

Some DNS servers I've run have used 1% CPU time on a
Pentium-166, some have used >50% on a P3-1200.

http://www.coker.com.au/selinux/   My NSA Security
Enhanced Linux packages
http://www.coker.com.au/bonnie++/  Bonnie++ hard
drive benchmark
http://www.coker.com.au/postal/    Postal SMTP/POP
http://www.coker.com.au/~russell/  My home page

Do you Yahoo!?
The New Yahoo! Search - Faster. Easier. Bingo.

Reply to: