Numpy: The numpy documentation is down and unreachable

Known issue, people are on it.

Any ETA on this? Any CI docs builds that utilize intersphinx from SciPy are failing

Not sure. Cc'ing @dpinte on this issue.

The server has been completely migrated to a new clean VM overnight. There are some final little glitches related to undocumented features that need to be migrated and we will be good to go. ETA is end of day today!

Awesome, thanks for the update Didrik

Would it be possible to use the CloudFlare always-online cache? This is not the first time I've noticed scipy.org is down.

Is there any alternative to access docs?

Can we build the docs on a mac?

Is there any alternative to access docs?

http://scipy.github.io/devdocs/ and http://devdocs.io/numpy~1.12/ ?

Suggestion to move docs hosting to github : https://mail.scipy.org/pipermail/numpy-discussion/2017-March/076578.html

The key issue is not hosting in this case but the documentation of what's been stored, where and how. We're working with machines that have been installed many years ago, have seen multiple maintainers touching configuration with various level of expertise, etc. It is a nice example of what entropy is!

Once everything will be back online, we can investigate any other hosting ideas. For the future, the real thing that needs to be addressed is the documentation of the infrastructure and ensuring that a policy existing for maintaining that infrastructure.

The scipy.org site is back online but there is a 500 for docs.scipy.org and when accessing www.scipy.org directly.

Still can't access it. Can anyone at least attach here objects.inv? Our documentations depends on it and we don't have cached version now.

Thanks to all the nice sysadmin working on that, good luck! you are the best :1st_place_medal:

docs.scipy.org is back online!

The performance is not there yet but the pages are served!

Intershinx is still not working:

sphinx.errors.SphinxWarning: WARNING: intersphinx inventory 'http://docs.scipy.org/doc/numpy/objects.inv' not fetchable due to <class 'requests.exceptions.ConnectionError'>: HTTPSConnectionPool(host='docs.scipy.org', port=443): Max retries exceeded with url: /doc/numpy/objects.inv (Caused by NewConnectionError('<requests.packages.urllib3.connection.VerifiedHTTPSConnection object at 0x2ab30b6ebac8>: Failed to establish a new connection: [Errno 110] Connection timed out',))

@bryevdv investigating the issue. I see there is something wrong:

$ wget http://docs.scipy.org/doc/numpy/objects.inv
--2017-03-15 17:48:01--  http://docs.scipy.org/doc/numpy/objects.inv
Resolving docs.scipy.org (docs.scipy.org)... 50.17.248.72
Connecting to docs.scipy.org (docs.scipy.org)|50.17.248.72|:80... connected.
HTTP request sent, awaiting response... 301 Moved Permanently
Location: https://docs.scipy.org/doc/numpy/objects.inv [following]
--2017-03-15 17:48:01--  https://docs.scipy.org/doc/numpy/objects.inv
Connecting to docs.scipy.org (docs.scipy.org)|50.17.248.72|:443... connected.
ERROR: The certificate of ‘docs.scipy.org’ is not trusted.
ERROR: The certificate of ‘docs.scipy.org’ hasn't got a known issuer.

Independently of the issue, could you switch to use https in place of http?

You might want to append the intermediate CA cert to the the server's cert. Most browsers will look for the intermediate cert but not all client libraries.

Certificate chain
 0 s:/C=US/ST=Texas/L=Austin/O=Enthought, Inc./OU=IT/CN=*.scipy.org
   i:/C=US/O=DigiCert Inc/OU=www.digicert.com/CN=DigiCert SHA2 High Assurance Server CA

The Certificate matching the issuer above: /C=US/O=DigiCert Inc/OU=www.digicert.com/CN=DigiCert SHA2 High Assurance Server CA

@bryevdv can you check now? Apache should be much happier than 10' ago

I'm still seeing this on Travis:

WARNING: intersphinx inventory 'https://docs.scipy.org/doc/numpy/objects.inv' not fetchable due to <class 'requests.exceptions.SSLError'>: [SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed (_ssl.c:749)

Actually here is the Qualys Labs SSL report saying the same thing: https://www.ssllabs.com/ssltest/analyze.html?d=docs.scipy.org

Look under "chain issues" and "Certification paths":

SSL will still work under Firefox, chrome, safari, etc because they will download that certificate on their own, but client libraries like requests probably will not.

Great job everybody!

@dpinte still failing:

https://travis-ci.org/bokeh/bokeh/jobs/210962890#L1674-L1755

@bryevdv now it fails with a certificate issue, which make sense. We will bundle the intermediate cert and that will fix the issue.

@bryevdv the intermediate certificate has been added locally. You should be able to download the file as expected

Thanks, working now

The server still seems really slow to access for me...

@dopplershift can you provide timings? We're continuing to monitor it and still tweaking some Apache configurations.

They seem more reasonable now, though it's still taking Safari 11 seconds to load the page. At least my sphinx build isn't timing out now, though.

11 seconds is a lot. Which page do you try to load?

Just https://docs.scipy.org/

When I tried just now, though, things were much better.

Things are perfectly fine now, from my tests.
Good job everybody :+1: !

FYI I'm seeing timeouts on intersphinx accessing objects.inv which is causing our builds to fail again. Navigating to spicy.org by hand, just tool 8-10 seconds to load

Just took about 25 seconds to load here

Confirmed. Investigating. The machine load is low. More than likely some tweaks for the Apache configuration

Not sure if I should post here or open anew, but seeing problems again.

same, docs.scipy.org is down and CI builds that rely on intersphinx are failing.

ping @dpinte @rgommers

@bryevdv looking at it. AWS console reported it has been unreachable since 31'. Restarting it

@bryevdv we should be back online. There seem to be some hardware issue with the AWS VM. Restarted and fully updated now

Thanks!