not seeing some brid.gy source URLs

[snarfed]

not sure of the root cause yet, but webmention.io returns 400 'The source URI does not exist' on at least some brid.gy webmentions. here's an example:

source: https://www.brid.gy/like/twitter/aaronpk/418251719303495680/14447132
target: http://ttk.me/t4Tq3
full log: https://www.brid.gy/log?start_time=1388621265&key=aglzfmJyaWQtZ3lySwsSCFJlc3BvbnNlIj10YWc6dHdpdHRlci5jb20sMjAxMzo0MTgyNTE3MTkzMDM0OTU2ODBfZmF2b3JpdGVkX2J5XzE0NDQ3MTMyDA

a couple guesses:

• does webmention.io cache source fetches? if so, maybe brid.gy choked the first time webmention.io asked for it, and it hasn't retried?
• the brid.gy URL is served over SSL but has mixed content due to the profile picture. (i'll fix that soon.) maybe webmention.io is picky?

[snarfed]

just fyi, i fixed the SSL mixed mode error, but the webmention is still failing, so it wasn't that.

[snarfed]

the webmention finally went through. not sure what changed. meh!

https://www.brid.gy/log?start_time=1388736588&key=aglzfmJyaWQtZ3lySwsSCFJlc3BvbnNlIj10YWc6dHdpdHRlci5jb20sMjAxMzo0MTgyNTE3MTkzMDM0OTU2ODBfZmF2b3JpdGVkX2J5XzE0NDQ3MTMyDA

[snarfed]

happening again, fwiw: https://www.brid.gy/#twitter-t . the accepted one may have been a fluke.

[aaronpk]

This is bizarre! A source_not_found error means webmention.io couldn't retrieve the source URL. (This is before it even looks for the link to target.) Here is the source where it returns that error.

I have been able to reproduce this by manually creating the scraper object and making a request for the page. It seems that the first time I make the request it's fine, and any later requests to www.brid.gy fail with OpenSSL::SSL::SSLError: SSL_connect SYSCALL returned=5 errno=0 state=SSLv3 read server hello A

The only ref to this I was able to find is rubygems/rubygems#522 (comment) which says "It seems the "hello A" errors also occur on request timeouts and slow links."

At the very least I should update webmention.io to respond with SSL failures with a separate code so you at least get a little better idea of what's wrong.

[snarfed]

huh! bizarre indeed. thanks for investigating. for my part, i have some low hanging fruit for speeding up those requests on brid.gy's side. i'll do some benchmarks and try those.

[snarfed]

fwiw, i'm seeing median client side response times of around .5s, with spikes up to .9s. server side is around 150ms. not great, but not horrible. maybe you can increase the timeout? i'll still plan to optimize on my end.

$ foreach i (`seq 0 10`)
foreach? time curl -so /dev/null https://www.brid.gy/like/twitter/t/419207701747761152/284933
foreach? end
0.033u 0.005s 0:00.49 6.1%  0+0k 0+0io 0pf+0w
0.032u 0.005s 0:00.47 6.3%  0+0k 0+0io 0pf+0w
0.034u 0.006s 0:00.53 5.6%  0+0k 0+0io 0pf+0w
0.034u 0.005s 0:00.61 4.9%  0+0k 0+0io 0pf+0w
0.034u 0.006s 0:00.44 6.8%  0+0k 0+0io 0pf+0w
0.034u 0.006s 0:00.47 6.3%  0+0k 0+0io 0pf+0w
0.035u 0.006s 0:00.44 6.8%  0+0k 0+0io 0pf+0w
0.033u 0.006s 0:00.49 6.1%  0+0k 0+0io 0pf+0w
0.034u 0.006s 0:00.89 3.3%  0+0k 0+0io 0pf+0w
0.039u 0.006s 0:00.44 6.8%  0+0k 0+0io 0pf+0w
0.033u 0.006s 0:00.45 6.6%  0+0k 0+0io 0pf+0w

[aaronpk]

I think it's actually a timeout in negotiating the SSL connection, before it even tries to make the HTTP request.

[aaronpk]

We tracked this down to what appears to be an issue with the SSL cert on www.brid.gy. I'll leave this ticket open until @snarfed installs the updated cert and we verify that it's working.

I also created a new issue #15 to return better errors to help debug this kind of thing in the future.

[snarfed]

this turned out to be an SNI issue. updating one of the ruby gems fixed it, so our debugging was cut short, but we captured some info in IRC and on the wiki.