Re: [Savannah-hackers-public] Email on fencepost hosed?

[Bob Proulx]

Hi Eli,

Eli Zaretskii wrote:
> Bob Proulx wrote:
> > I just sent a test to myself on fencepost and it is working fine for
> > me.
> 
> Later messages indicate that you did this after the problem was
> already solved.

Yes.  By that time things were already fixed.  But that is what I
tested and saw and so reported it thusly.

> > > and SMTP connections to fencepost are refused.
> > 
> > That is normal.  Mail for @gnu.org addresses are received by eggs.
...
> I'm not an expert, but my mail is set up according to instructions
> here:
>   https://www.fsf.org/about/systems/sending-mail-via-fencepost
> which explicitly say:
>   SMTP server: fencepost.gnu.org

Ah...  Understanding dawns on me.  I made a bad assumption on my part.
I was assuming you were trying to *receive* mail on fencepost.
Because when I read smtp into fencepost I jumped to the reception of
incoming email.  But you are also using it to send email out *through*
fencepost.  I had forgotten about that path through things.  Sorry.

> So I'm not sure I understand the "normal" part.  What did I miss?

Anyone who has an account on fencepost may send mail from there and
may receive and read mail to there without any special configuration.
In my mind receiving mail that is "normal" because it doesn't have any
configuration.  It just comes with having a fencepost account.  (Just
log in and start 'emacs' and go.  Or 'mutt'.)  Plus anyone can send
mail to your fencepost address.  They just send email to your address
and they don't know they are doing anything.  But for the general
public when they send mail to your fencepost account all of their mail
goes through the MX records pointing to eggs.  For most of the world
eggs receives the email, has some frontend anti-spam, then relays it
along to fencepost.  Fencepost itself doesn't interact with the public.

But you were talking about sending mail through fencepost remotely as
an account holder.  Which on fencepost is done using SMTP as you
state.  Which means that a mail-transfer-agent must be running on port
25 smtp on fencepost for you to interact.  But this is only for people
who both have an account on fencepost and have set up their
configuration to send mail through fencepost.  Which is a smaller
subset of the population.  And one I had forgotten about.

> > > Could someone please look into this, or at least update
> > > https://pumprock.net/fsfstatus with information about what's going on?
> > 
> > None of us here have anything to do with admining fencepost, lists, or
> > the pumprock fsfstatus page.  That is only the FSF admins.  And of
> > course the way to reach them is with a mail to address@hidden  They
> > keep a tight lock on access to all of those things.  So even if things
> > were really messed up and broken all we could do is comiserate with
> > each other about it.
> 
> Please trust me that I started by sending email to address@hidden as
> soon as I saw the problem.  I posted here after seeing that nothing
> happened for several hours after my mail.  Without a good
> understanding of the gnu.org email delivery, I couldn't know whether
> my email at all reached the addressee (and I specifically said that in
> my message posted here).

Yes.  I feel your frustration coming through your words.  I know that
John Sullivan is often pushing the FSF sysadmins to be more
transparent in their operations and to keep notifications current.
But unfortunately people are people and in the heat of the moment
those things often get forgotten.  Perhaps the best thing we can do is
to keep the feedback communication active so that they know that the
notificates are both really appreciated when received and really
missed when they are not.  Eventually it should become part of the
culture.  Until then we just must keep reminding them.

> If there's nothing else to do in these cases but simply wait for
> someone to detect the problem, then I think we have a weakness in the
> setup that should perhaps get some attention.

It appears to me from reading the irc chatter that there was supposed
to be an automated notification through Nagios back to the admins.
But that appears to have failed.  The real follow-up is if that check
has been debugged and repaired?  I will follow-up on that item so that
it isn't forgotten.

> What else could I have done to speed up the correction of the
> problem?  (Or maybe there are already instructions how to behave in
> these situations, and I just missed them?)

Let me apologize by the tone of my message.  I was trying to report
facts.  But obviously some tone crept in.  Sorry.  Because often
people are asking why can't we do something and so we try to make it
clear that due to the division of labor we have one area and they have
a different area and all we can do is redirect people to the parties
responsible for that area.

You already knew about the pumprock status page.  That is good.  I
think it is often out of sight and out of mind though and too often
needs prodding to get it updated.  I am hoping that changes as the
culture of updating that becomes more ingrained.  I am hoping.

Fencepost is a smaller subsection of the entire universe in Boston.
The admins don't routinely use it themselves (I don't think) and
therefore if there are problems with it then they will need to be told
about it.  I don't think it would get noticed otherwise.  So notifying
them is definitely the best thing to do.

Karl added the address@hidden address as well.  That was a good
additional bit of information.  In this case because mail to the
@gnu.org addres should have been working okay (doesn't involve
fencepost) so either of those should get you in contact with the FSF
admins directly.  Then some patience is needed as there are a lot of
tickets entering the queues.

For incoming mail if eggs is down then that is a really, really big
deal and I am sure it will get noticed by multiple methods.  Because
pretty much everything incoming goes through eggs.  Certainly all of
the mailing list traffic.  When it is down I brace for the tsunami
wave of backlog that will start up again once it is brought back
online.

Other than this I don't know of anything better that can be done.

Bob