draeder
commented
3 years ago Adding the following seems to speed up failed/timedout socket connection responses, per https://stackoverflow.com/a/66215639/1551027, along with keepAliveMsecs: timeout from the node documentation:
handler.Agent({keepAlive:true, keepAliveMsecs: timeout, timeout: timeout})And also this per https://stackoverflow.com/a/53362571/1551027
maxSockets: 20Finally putting it all together:
var options = urlToHttpOptions(new URL(url), handler.Agent({keepAlive:true, keepAliveMsecs: timeout, timeout: timeout, maxSockets:20}));I tested it on this line:
Which seemed to resolve almost all of the issues I was having with ETIMEDOUT at TCPConnectWrap.afterConnect [as oncomplete], but I'm stuck on one or two straggling problematic tracker servers that still cause this timeout exception and aren't honoring the timeout settings in the code.
theobarrague
Orignal message from draeder in this issue
It looks like it's working for the most part. What I'm doing is building a package that uses fake-bittorrent-tracker to test the list of trackers from trackerslist. First I simply ping them and remove dead hosts. Next I use fake-bittorrent-tracker as designed to upload/download a fake torrent to the remaining hosts. I can see that the timeout is working for the majority of remaining trackers in my list.
But for some of the trackers, the timeout doesn't seem to work. This might be due to lower level networking failures, like SSL failures when trying to open the socket, or general socket failures. Here is one example where fake-bittorrent-client took well over 30 seconds to finally timeout, even with the updated code and the timeout set to 1500ms:
That particular tracker URL seems to have a variety of problems, sometimes it 404s, sometimes it has a socket reset, and other times the timeout above; so it's a good URL to test with.