Re: [Savannah-hackers-public] Git on Savannah rejects git:// protocol?

[Bob Proulx]

Eli Zaretskii wrote:
> I get this:
> 
>   ~/test-clone$ git clone git://git.savannah.gnu.org/git/emacs.git
>   Cloning into 'emacs'...
>   fatal: unable to connect to git.savannah.gnu.org:
>   git.savannah.gnu.org[0: 2001:470:142::168]: errno=Connection refused
>   git.savannah.gnu.org[1: 209.51.188.168]: errno=Connection refused

If connections are being refused then the git-daemon is unable to
service incoming on port 9418 for some reason.  Most likely the system
is overloaded.

> But:
> 
>   ~/test-clone$ git clone https://git.savannah.gnu.org/git/emacs.git
>   Cloning into 'emacs'...
>   remote: Counting objects: 968472, done.
>   remote: Compressing objects: 100% (177822/177822), done.
>   remote: Total 968472 (delta 795603), reused 960310 (delta 789344)
>   Receiving objects: 100% (968472/968472), 349.50 MiB | 7.58 MiB/s, done.
>   Resolving deltas: 100% (795603/795603), done.
>   Checking out files: 100% (4575/4575), done.

Nginx is a pre-forking daemon.  git-daemon is not.  (Or at least not
as configured if that is possible to do so.)

> Does this mean Savannah no longer supports the git:// protocol?

Definitely not.  And git:// is working great now, several hours later
after your problem report.

> Or is this an unintended consequence of some change in the
> configuration lately?

Possibly.  Or possibly it would have happened on the old server too.
Don't know.  Because it is a combination of the server and the
interaction with the Internet.  The first is at least knowable but the
second is always changing and often hostile.

Another thing that I note is that for whatever reason there seems to
be more Internet abuse happening in the hours from 3am to 6am
(local'ish time, the FSF admins are US/Eastern timezone and I am
US/Mountain timezone two hours later) than at other times of the day.
This is subjective but it always seems like problem reports of
problems will be occurring at that time of day more so than during US
daylight hours.

This has a more proportionate effect on our overseas free software
friends who operate mostly during that time.  Other than being awake
during that time myself it makes it hard to diagnose what happens then
when it recovers and is fine some hours later.  I have this belief
that more fail2ban automation to detect abuse and to automatically
mitigate it is needed.

Hmm...  Did I port over all of the fail2ban and other dynamic
mitigation from the old server to the new?  I did not!  That is almost
certainly a contributor to the problem seen last night.  It's running
in a default configuration and not our very customized one.

I'll read and respond to your next message.

Bob