api.rubygems.org Connection reset by peer - SSL_connect

[mphilipps]

hi,
Since about 5 days ago 1 out of ~15 https connections to api.rubygems.org fails.

gem list -V --debug --remote '\Ayard\z'
NOTE: Debugging mode prints all exceptions even when rescued
Exception `LoadError' at /opt/puppetlabs/puppet/lib/ruby/2.4.0/resolv.rb:171 - LoadError

*** REMOTE GEMS ***

Exception Errno::EEXIST' at /opt/puppetlabs/puppet/lib/ruby/2.4.0/fileutils.rb:230 - File exists @ dir_s_mkdir - /root/.gem/specs/api.rubygems.org%443 Exception Errno::ECONNRESET' at /opt/puppetlabs/puppet/lib/ruby/2.4.0/net/protocol.rb:44 - Connection reset by peer - SSL_connect
Exception Errno::ECONNRESET' at /opt/puppetlabs/puppet/lib/ruby/2.4.0/net/http.rb:964 - Connection reset by peer - SSL_connect Exception Gem::RemoteFetcher::FetchError' at /opt/puppetlabs/puppet/lib/ruby/2.4.0/rubygems/remote_fetcher.rb:311 - Errno::ECONNRESET: Connection reset by peer - SSL_connect (https://api.rubygems.org/latest_specs.4.8.gz)
Exception `Gem::RemoteFetcher::FetchError' at /opt/puppetlabs/puppet/lib/ruby/2.4.0/rubygems/spec_fetcher.rb:265 - Errno::ECONNRESET: Connection reset by peer - SSL_connect (https://api.rubygems.org/latest_specs.4.8.gz)

The server in question is behind a NAT, but not a popular one (aws/google/azure) and is also the only server behind it that is regularly connecting to rubygems.org

[sonalkr132]

Hi @mphilipps

Can you please follow this troubleshooting guide let us know about results?

[mphilipps]

hi,
Not using bundler here so that fails as expected:

Other than that it looks like this:

Bundler connection to rubygems.org: failed ❌ (uninitialized constant Bundler)
RubyGems connection to rubygems.org: success ✅
Ruby net/http connection to rubygems.org: success ✅

except when it fails in that case it shows:

Bundler connection to rubygems.org: failed ❌ (uninitialized constant Bundler)
RubyGems connection to rubygems.org: failed ❌ (Errno::ECONNRESET: Connection reset by peer - SSL_connect (https://rubygems.org))
Ruby net/http connection to rubygems.org: success ✅

With that it seems the behaviours is exactly the same as in my original report.

Skiming over the rest of the page:

• ntp is installed, running and the time is correct.
• I don't think messing with the installed versions/certificates is going to fix it, given that is works 95% of the time.

It feels more like a load balancing issue on rubygems.org side.

[indirect]

@mphilipps our frontend servers are run by Fastly. I recommend visiting http://fastly-debug.com and using the widget in the bottom right to open a ticket about the connection resets you are seeing. 👍

[sonalkr132]

hi @mphilipps are you still facing this issue?

[mphilipps]

Not at the moment, after a couple of days it disappeared and then resurfaced like once or twice.
I am closing this as it has been fine for the last few weeks.