Re: Blocking Calls and dnsserver avg service

Hi Torsten

> 1. Why does squid report:
>
> "Blocking calls to gethostbyname(): 4"
>
> but on the other side
>
> "dnsservers use histogram:
> dnsserver #1: 160
> dnsserver #2: 11
> dnsserver #3: 0
> dnsserver #4: 0
> dnsserver #5: 0
> dnsserver #6: 0
> dnsserver #7: 0
> dnsserver #8: 0
> dnsserver #9: 0
> dnsserver #10: 0"
What operating system are you running it on? Which version of squid are you
using?

> 2. Is a "dnsserver avg service time" of 2714 msec
> normal for a good internet connection (2 MBit/s) ??
It depends on the type of link - if you have a fibre link it is not doing
as it is supposed to, but if you have a sattelite system you are probably
sending about 2 requests, one to the root servers, and then one to the
authoritive servers. At about 500ms per trip it will take about 2000 ms

(ours generally does)

> Should we even run a separate named on the squid machine ?
Nope - if your name server has enough memory and CPU power it is
better to have a loaded dns server than a unloaded one - the DNS server
caches responses and answers them directly from ram for a good few hours
(normally) thus your responses are much faster than if you connect via an
international link.

> Background: We are experiencing strong delays for our customers in times
> of heavy load. But where can I find out, if the delay is caused by our
> parent squids, by our squid itself, or by slow hardware ...?

Try running "iostat" or "vmstat" when your server starts slowing down. This
should give you some idea as to how much disk activity there is.

In your "acl"'s are you limiting people to having to come from a specific
domain or a specific IP address range? Limiting the number of DNS queries that
you make would speed things up.

There is a useful program called echoping, from
ftp://ftp.pasteur.fr/pub/Network/echoping

Which my help you too.

        Oskar

=============================================
'experience made art, but inexperience luck.'
=============================================
Received on Tue Jan 14 1997 - 06:26:51 MST