On 10/05/07, <b class="gmail_sendername">bert hubert</b> <<a href="mailto:firstname.lastname@example.org">email@example.com</a>> wrote:<div><span class="gmail_quote"></span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
On Thu, May 10, 2007 at 02:22:42PM +1000, Amos Shapira wrote:<br>> 1. A Geo backend which will pick a CNAME record based on the IP from which<br>> the query came.<br>> 2. A database backend (MySQL/PostgresSQL) which will resolve the name that
<br>> the CNAME points to into an IP address.<br>><br>> I wonder if that I wrote until here makes sense.<br><br>Very much so!<br><br>> Now the twist is that I was thinking of having a specialised program which
<br>> probes all the available servers, checks that they are still alive and well,<br>> and updates the records of the database back-end whenever a server's<br>> availablity changes.<br><br>This is one easy way of doing things, and it would work fine.
<br><br>> Is this a possible plan? How do others setup such networks?<br>> e.g. I read the Wikipedia uses PowerDNS with GeoIP to forward users to<br>> nearer servers, but does it also check whether the node or cluster it
<br>> forwards the user to is available at all?<br><br>There are no special checks for the Wikipedia. Some other PowerDNS users<br>have used a special 'pipe' backend to do realtime checks of server<br>availability, but changing the database on server availability changes works
<br>as well.<br><br>Good luck!</blockquote><div><br>Thanks very much for the feedback. It helps a lot to my trust in my plan.<br><br>About the pipe backend - it sounds awfully slow to do such "realtime" checks (at least in my case - the servers could be halfway around the world). If I read the docs correctly then PDNS is smart enough to cache the results so not all clients will be hit by a check, only the first one - is this correct?