Re: [Savannah-hackers-public] Getting git errors with http URLs

[Bob Proulx]

Hi Arnold,

> $ git pull
> fatal: unable to access 'https://git.savannah.gnu.org/r/gawk.git/': The 
> requested URL returned error: 502
> 
> Using a git://... URL works with no problem.
> 
> What's the scoop?

There has been "an abuse" (maybe not so far as an attack, I think an
attack would take us completely offline) going on the past few days
that is draining resources.  A large set of IP addresses is walking
through every web browsable link for every project.  This includes
cgit, viewvc, loggerhead, everything.  This has been going on since
the 14th.  Perhaps someone is trying to mirror everything in an
abusive way?  I don't know.  I have been watching it and trying to
mitigate it.  But there are a lot of IP addresses.

The result is that it starves the system of resources.  It consumes
processes.  It consumes memory.  None of the failures you are seeing
are hard failures.  Another attempt and they might get through the
queue ahead of the abusers.

The ssh:// protocol and the git:// protocol uses a different set of
processes and have their own process limits.  ssh being the most
permissive since that is a fully authenticated protocol.  I think
people using ssh:// are not noticing a significant problem.  Since
https shares with viewvc and cgit it is getting starved along with
them.

I have been traveling around myself since Thursday[1] which prevented
me from having enough time to really dig into things.  But I am back
and looking at what might be done.  However with distributed attacks
such as this they can be quite difficult to avoid.  I started looking
in depth last night but unfortunately my first attemps to mitigate
this didn't help significantly.

I am digging into it in detail now.

Bob

[1] I left for Pogosa Springs on Thursday in my C140.  But then had a
fuel starvation problem with my left fuel tank!  Vents are in the fuel
caps.  If they get clogged up then no air goes in.  If no air goes in
then no fuel comes out.  No fuel comes out and the engine doesn't keep
running.  I landed at Boulder (just happened to be near me at that
time) to sort things out.  Thought I did.  It is somewhat hard to tell
when you are putting your mouth on the spout of the fuel cap and
trying to pull air through to see if the vent is okay or not.  (One of
these days I want to build a fuel cap vent tester specifically for
this purpose.)

Took off to continue the flight.  But then the problem with that tank
persisted.  I had to turn around and fly back home on the other tank.
Redundancy is a good thing.  Then much scrambling around with the
mechanics trying to resolve the problem.  But when you have a 1946
airplane the parts are not always immediately available.  This one is
going to need some digging.  I really want to replace the fuel cap but
getting an appropriate one isn't always available.  This is where the
vintage type clubs are great resources.  It was late before I got back
to my house.  With the problem still not completely resolved.  Needs
more work.

But I still needed to get to Pogasa Springs for a combined work trip
and personal trip.  Therefore on Friday I swapped over to a different
airplane.  I took my C182 down to Pogasa Springs instead.  Sure the
C182 is the better mountain airplane by a lot but the C140 is also
just a joy to operate.  In any case it was uneventful flying down and
back.  Flew back Saturday.  Started looking at the Savannah resource
starvation problems on Saturday night but was ineffective.