Ulysses
03-02-2005, 12:10 AM
G'day,
I have encountered the following negative argument for running DNS servers from the same box, by a competitor, who constantly uses this argument to scare customers away from other hosts, presumably to him. Is it true? Here it is:
The whole reason why you are FORCED to specify 2 name servers is to prevent DNS cache records from expiring and then caching a DNS failure.
The way it works is this:
An ISPs TTL (time to live) for a particular domain will expire, the ISPs DNS server then queries the name servers set for the domain to obtain an updated TTL and A record for the domain.
X ISPs DNS server will query ns1.something.com (whatever is specified for the domain in question). If that times out, it will query ns2.something.com and so on until it runs out of name servers to query.
If all of the name servers specified for a domain time out, the domain will be cached as non-responsive (poisoned) on X ISPs DNS server so that bandwidth / processing power is not wasted continually trying to query failed name servers.
When the name servers finally do come back online, the ISP still won’t query those name servers until the poisoned cache expires – a time which is defined by each ISP.
The net result is that the name servers and host server (in this case all on the same box) can be online, while ISPs will return poisoned cache results to the client – so the site will still look down even though the host is up.
Running DNS servers even from the same network is a VERY bad idea. Running DNS servers from the same box is…suicide.Is any of this correct? If it is, is it really a "suicidal" problem?
No hurry guys, but I'd really like to get around this argument - and "outsmart" this bloke. :cool:
Michael
I have encountered the following negative argument for running DNS servers from the same box, by a competitor, who constantly uses this argument to scare customers away from other hosts, presumably to him. Is it true? Here it is:
The whole reason why you are FORCED to specify 2 name servers is to prevent DNS cache records from expiring and then caching a DNS failure.
The way it works is this:
An ISPs TTL (time to live) for a particular domain will expire, the ISPs DNS server then queries the name servers set for the domain to obtain an updated TTL and A record for the domain.
X ISPs DNS server will query ns1.something.com (whatever is specified for the domain in question). If that times out, it will query ns2.something.com and so on until it runs out of name servers to query.
If all of the name servers specified for a domain time out, the domain will be cached as non-responsive (poisoned) on X ISPs DNS server so that bandwidth / processing power is not wasted continually trying to query failed name servers.
When the name servers finally do come back online, the ISP still won’t query those name servers until the poisoned cache expires – a time which is defined by each ISP.
The net result is that the name servers and host server (in this case all on the same box) can be online, while ISPs will return poisoned cache results to the client – so the site will still look down even though the host is up.
Running DNS servers even from the same network is a VERY bad idea. Running DNS servers from the same box is…suicide.Is any of this correct? If it is, is it really a "suicidal" problem?
No hurry guys, but I'd really like to get around this argument - and "outsmart" this bloke. :cool:
Michael