bug#30290: guix-daemon slows to a crawl when a substitute server is offline

[Maxim Cournoyer]

Hi,

And sorry for failing to produce a reply earlier :-).

zimoun <zimon.toutoune@gmail.com> writes:

[...]

>>> From: Maxim Cournoyer <maxim.cournoyer@gmail.com>
>>> Date: Fri, 18 Dec 2020 22:04:04 -0500 (24 weeks, 4 days, 18 hours ago)
>>>
>>> I'm not sure if the recent offloading work that Mathieu did touched that
>>> topic.  I'd need to test the scenario.  Perhaps a system test would be
>>> useful.
>>> ----------
>>>
>>> From: Ludovic Courtès <ludo@gnu.org>
>>> Date: Tue, 22 Dec 2020 16:16:08 +0100
>>> Date: Tue, 22 Dec 2020 16:16:08 +0100 (24 weeks, 1 day, 6 hours ago)
>>>
>>> Is it still a problem?  Commit 4f5234be0378368e6af25925db46612838d25e58
>>> (Nov. 2019) added a table of unreachable hosts.  That way, a ‘guix
>>> substitute --query’ process won’t retry connections to an unreachable
>>> host.
>>> ----------
>>>
>>> From: Efraim Flashner <efraim@flashner.co.il>
>>> Date: Mon, 28 Dec 2020 14:19:02 +0200
>>> Date: Mon, 28 Dec 2020 14:19:02 +0200 (23 weeks, 2 days, 9 hours ago)
>>>
>>> Occasionally my internet drops itself, and I find I'm left forever
>>> waiting for a timeout to see what sources I have cached locally.
>>> ----------
>>
>> What is the current stats of this bug?   Is it still happening with the
>> recent improvements of Cuirass?
>
> After reading all this, I think this bug can be closed.  WDYT?

Were you able to replay a scenario in which a substitute server is made
unreachable?  That's the information that I'd like to have/see before
closing.  I don't come across unreachable substitute servers often, and
can't think of a way to easily test this.

I could make it hang by dropping the input/output connections with
iptables to a remote guix publish server, but then SSH also hangs, so
perhaps that's expected.

I'll try to configure a couple local machines to act as publish servers,
and disconnect them from the network to see what happens.

Thanks,

Maxim