<div dir="ltr">Thanks everyone who helped make this happen.<div><br></div><div>From my perspective* I believe the ssh upload mechanism was added to address security issues around the basic-auth-over-http method used historically. Now uploads *may* be done over https, and those using the ssh method can move over to using twine or pip upload, I think that it's reasonable to discontinue support for ssh uploads.</div>
<div><br></div><div><br></div><div> Richard</div><div><br></div><div>* as the guy who will be hassled if its loss is noticed ;)</div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On 26 January 2014 10:38, Donald Stufft <span dir="ltr"><<a href="mailto:donald@stufft.io" target="_blank">donald@stufft.io</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Today (Sat Jan 25, 2014) the Infrastructure team has migrated PyPI to new<br>
infrastructure.<br>
<br>
The old infrastructure was:<br>
<br>
- a single database server managed by OSUOSL<br>
- a pair of load balancers shared by all of the <a href="http://python.org" target="_blank">python.org</a> services hosted on<br>
OSUOSL<br>
- a single backend VM that served as everything else for PyPI<br>
<br>
The VM that was acting as the backend server from PyPI was partially hand<br>
configured and partially setup to be managed by chef. Additionally it had an<br>
issue that caused it to kernel panic every so often which had been the cause of<br>
a number of downtimes in the last few months. Because it was primarily<br>
configured and administered by hand and because the way it was set up it was<br>
not feasible to have any sort of failover or spare.<br>
<br>
The new infrastructure is:<br>
<br>
- 2 Web VMs<br>
- 2 Database servers in Master/Slave Configuration<br>
- 2 PgPool Servers pooling connections to the database servers and load<br>
balancing reads across them.<br>
- 2 GlusterFS servers backed by Cloud Block Storage acting as the file storage<br>
for package and package docs<br>
- 1 metrics server to handle updating the download counts as they come in from<br>
Fastly<br>
<br>
All of the VMs are hosted on Rackspace’s Public Cloud and have their<br>
configuration completely controlled and managed using Salt. Going forward this<br>
will allow us to easily scale out as required or kill malfunctioning servers<br>
and spin up new ones easily. Additionally the setup has been setup so that<br>
where possible there is two servers performing the same role, ideally in an<br>
Active/Active configuration but at least in a Master/Slave configuration. This<br>
should allow PyPI to be far more stable moving forward and make downtimes much<br>
easier to recover from.<br>
<br>
The services are still fronted by Fasty’s CDN and in the new infrastructure<br>
we’ve removed our load balancer and have replaced it by having Fastly handle<br>
the load balancing for us. Additionally we’ve recently setup a static mirror of<br>
PyPI that is updated once every minute. This is hosted on Rackspace cloud as<br>
well but in a separate data center from the rest of PyPI. Fastly is configured<br>
to fall back to this static mirror in the case that neither of the two web<br>
heads are functioning. This should ensure that even in the event of a<br>
catastrophic failure of the PyPI service that the bulk of package installations<br>
should hopefully remain working.<br>
<br>
The bad news, (and the “Breakage” from the subject) is that while the new<br>
infrastructure was being planned out, built, and migrated to the “pypissh”<br>
package was forgotten. The pypissh package is an alternative way to upload<br>
packages to PyPI however it is very difficult, because of the way it works, to<br>
provide HA support for it as we’ve set up for everything else. We don’t have<br>
any numbers for how many people are actively using this package but looking<br>
at a roughly 2 week chunk of time in PyPI’s download history, the pypissh<br>
package was downloaded 7 times by a browser, and 7 times by pip. All other<br>
downloads were caused by the mirroring system.<br>
<br>
As of right now pypissh is non functional and due to the difficulty in HAing<br>
and monitoring the current setup and because it is apparently has a very<br>
small set of users we would like to effectively kill off this particular<br>
service. Additionally the benefits of pypissh have been reduced now that PyPI<br>
is available over a TLS connection with a well trusted certificate. My question<br>
to you is, is this something that distutils-sig is willing to have happen? If<br>
we are to re-enable pypissh we’ll need to write a new solution to doing it that<br>
can be properly HA’d and we’d prefer to put our efforts into improving things<br>
for a much larger set of people.<br>
<br>
So yea, PyPI should be loads more stable and more reliable now.<br>
<br>
-----------------<br>
Donald Stufft<br>
PGP: 0x6E3CBCE93372DCFA // 7C6B 7C5D 5E2B 6356 A926 F04F 6E3C BCE9 3372 DCFA<br>
<br>
<br>_______________________________________________<br>
Distutils-SIG maillist - <a href="mailto:Distutils-SIG@python.org">Distutils-SIG@python.org</a><br>
<a href="https://mail.python.org/mailman/listinfo/distutils-sig" target="_blank">https://mail.python.org/mailman/listinfo/distutils-sig</a><br>
<br></blockquote></div><br></div>