Hi Alex,
Thanks for your detailed reply. Our replies are inlined
>
> If real DNS resolution does not work at all, then you will need to
> figure out why the configured DNS resolver is not getting Polygraph DNS
> queries or why Polygraph is not getting the responses. You can try a
> manual test with dig (bound to a polyclt IP alias!) to troubleshoot.
>
We checked with dig and we dont see any problem getting the names resolved.
> If real DNS resolution works, but generates too many errors/timeouts due
> to overload, I see two options (besides checking the network cable :-):
>
We checked. We also tested with iperf and we get the maximum
bandwidth between the dns and the polyclient. The tcpdump on the dns
shows the queries and response to and from the server. The dns daemon
also gets considerable load.
> 1) Hack Polygraph sources to avoid the resolution for all or a
> hard-coded set of domain names. The response(s) will be hard-coded. This
> should work fine for a one-time test.
>
How about using the library calls like gethostbyname() which can be
used to get the resolution through hosts files.
> 2) Teach Polygraph to use the PGL address map to resolve names without
> sending any DNS queries. This option is more work, but is the right
> solution, especially if the testing is not confined to a couple of
> weeks.
>
> In either case, your team can hack Polygraph sources or ask the Factory
> to do the work.
Thanks
--jag
Received on Thu Oct 26 13:08:27 2006
This archive was generated by hypermail 2.1.8 : Fri Oct 27 2006 - 12:00:06 MDT