Bug#321922: [pkg-fetchmail-maint] Bug#321922: fetchmail: DNS temporarily failure after suspend to disk

[Goswin von Brederlow]

Nico Golde <nion at gmx.net> writes:

> Hi,
> please quite correctly.
> * Wolfgang Kohnen <wolfgang.kohnen at lis.bremen.de> [2005-08-09 00:37]:
>> > * Wolfgang Kohnen [1]<wollie at lis.bremen.de> [2005-08-08 11:52]:
>> > > sorry if this is a known bug, BTS seems to be down or something...
>> > > 
>> > > After hibernating my notebook with software suspend v2 to disk for a
>> > > while (say 10 hours), fetchmail can'T resolve servers' hostnames and
>> > > comlains about temporarily DNS errors. A restart of the fetchmail daemon
>> > > fixes this behaviour.
>> > 
>> > [...]
>> > All the other programs are able to do this?
>> > Regards Nico
>> 
>> Hmm, I think I should give more background info: I use a patched 2.6
>> kernel with software suspend to disk. 
>
> Have you tried it with a vanilla kernel? Do you use special
> wlan drivers like ipw2100 or ipw2200?
>
>> The suspend script shuts down my
>> ehternet device before stopping all processes and then goes into suspend
>> state. After resuming and continuing the processes (there is a SIGCONT
>> signal I think) the suspend script itself resumes and brings up the
>> ethernet device again. Since I use dhcp for my eth0, the network
>> environment changes sometimes, i.e. different address, different default
>> gateway, different nameserver entry in resolv.conf and so on.
>
> shouldn't be a problem, fetchmail is not responsible for
> this. fetchmail doesn't read your resolv.conf for example.
>
>> So IMHO there are three possible pitfalls: The long suspend time which
>> creates a huge jump into the future from fetchmails point of view and
>> maybe something like a timer get confused. The "sudden" change of network.
>> The time between SIGCONT and net reestablishment. Just my two, no, three
>> cents. :-)
>
> I really don't think this is a fetchmail issue, fetchmail
> doesn't touch your network configuration, it just send the
> packages through your network device driver but doesn't
> handle it. I am pretty shure that this bug is on your site.
> I just tested it with Suspend to RAM/Disk I can not
> reproduce it.
> Regards Nico

It might just be libns trying to keep using the cached nameserver. But
nearly any dns using program that you started before hybernating
should get the same effect then.

MfG
        Goswin