Connection errors when there should be warnings

[QuintinWillison]

Reported by customer connecting to Ably "via the GraphQL ruby server's subscription".

https://graphql.pro/
https://graphql-ruby.org/subscriptions/ably_implementation.html

@mattheworiordan discussed the issue with the customer on Intercom (internal conversation). Specifically, Matt stated "You should have seen warnings, but not experienced errors".

Customer log output:

Ably::Exceptions::ConnectionError
and then SSL_connect SYSCALL returned=5 errno=0 state=unknown state (OpenSSL::SSL::SSLError) (code: 80000, base exception: Faraday::ConnectionFailed) -> see https://help.ably.io/error/80000 for help < SSL_connect SYSCALL returned=5 errno=0 state=unknown state (OpenSSL::SSL::SSLError)

Ruby stack track (full path redacted):

/.../ruby/2.6.0/gems/ably-rest-1.1.0/lib/submodules/ably-ruby/lib/ably/rest/client.rb:601:in `rescue in send_request'
/.../ruby/2.6.0/gems/ably-rest-1.1.0/lib/submodules/ably-ruby/lib/ably/rest/client.rb:541:in `send_request'
/.../ruby/2.6.0/gems/ably-rest-1.1.0/lib/submodules/ably-ruby/lib/ably/rest/client.rb:524:in `block in raw_request'
/.../ruby/2.6.0/gems/ably-rest-1.1.0/lib/submodules/ably-ruby/lib/ably/rest/client.rb:609:in `reauthorize_on_authorization_failure'
/.../ruby/2.6.0/gems/ably-rest-1.1.0/lib/submodules/ably-ruby/lib/ably/rest/client.rb:523:in `raw_request'
/.../ruby/2.6.0/gems/ably-rest-1.1.0/lib/submodules/ably-ruby/lib/ably/rest/client.rb:294:in `get'
/.../ruby/2.6.0/gems/ably-rest-1.1.0/lib/submodules/ably-ruby/lib/ably/rest/presence.rb:43:in `get'
/.../ruby/2.6.0/gems/graphql-pro-1.10.4/lib/graphql/pro/ably_subscriptions.rb:14:in `block in execute_all'
/.../ruby/2.6.0/gems/graphql-pro-1.10.4/lib/graphql/pro/subscriptions/redis_storage.rb:65:in `block in each_subscription_id'
/.../ruby/2.6.0/gems/graphql-pro-1.10.4/lib/graphql/pro/subscriptions/redis_storage.rb:60:in `each'
/.../ruby/2.6.0/gems/graphql-pro-1.10.4/lib/graphql/pro/subscriptions/redis_storage.rb:60:in `each_subscription_id'
/.../ruby/2.6.0/gems/graphql-pro-1.10.4/lib/graphql/pro/ably_subscriptions.rb:12:in `execute_all'
/.../ruby/2.6.0/gems/graphql-1.9.4/lib/graphql/subscriptions.rb:69:in `trigger'to which Matt observed:

Mm, odd, I would have expected in STDOUT to see some logs on retries etc. too

But the customer confirmed what they were seeing with:

3[31mERROR#033[0m Ably::Rest::Client - Request FAILED after 0 retry for get /channels/XXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX/presence {:limit=>100} (seq #JygtaQ, time elapsed 59.33s)

and:

We're using a proxy, and I believe we are now using persistent connections but ... I am not 100% sure.

[TheSmartnik]

The issue seems to be caused by the fact that proxy server client had used could only handle TLS 1.0 with cipher which no longer enabled by default in ruby (StackOverflow). However, we've changed our default client to typhoes (uses libcurl underneath) so the problem shouldn't happen again.

To sum up:

1. This error shouldn't be a warning. Ably client couldn't connect to the server and the fix in code would have been required
2. Since we've changed our middleware, the issue should no longer appear