Apache Perl/PHP/proxy DNS lookup failures
uncoolbob at gmail.com
Mon Dec 20 21:15:00 GMT 2010
On Mon, Dec 20, 2010 at 7:32 PM, Brian Manning <elspicyjack at gmail.com>wrote:
> On Mon, Dec 20, 2010 at 6:52 AM, Bob MacCallum <uncoolbob at gmail.com>
> > We (non-profit bioinformatics site maintainers) are seeing regular
> > intermittent DNS lookup failures from mod_perl, php and mod_proxy. Once
> > error appears it keeps occurring until you restart apache, even though on
> > the commandline the hosts can be looked up and contacted fine (e.g. with
> > ping). We've googled the error messages many times, and we're not the
> > to experience this [1,2] - for many it seems to happen when DHCP leases
> > renewed on home ADSL connections, and they are happy to restart Apache.
> > Looking at our resolv.conf timestamps, however, DHCP renewal doesn't seem
> > be the problem.
> Just curious, why do you need to do DNS lookups? If it's for logs,
> most log processing tools will do hostname lookups for you when they
> process your logs.
Thanks everyone for the hints - we'll follow those up.
The trivial uses for the lookups are where the front-end apache uses
ProxyPass directives to back end apaches/tomcats, and also for some
front-end PHP code to connect to the database host, all using hostnames.
(We have quite a mixture of technologies going on!)
Ironically, these hostnames are set up in /etc/hosts so that our
development, preview and production servers can each have different database
and other back end servers.
A more complex use is in one of our back ends - the Ensembl genome browser,
to which users can attach their own data sources from their own servers.
This is where the LWP hack seemed to work.
I believe we have hostname lookups switched off as recommended for logging,
More information about the london.pm