01:56:49  * DarkGodquit (Ping timeout: 255 seconds)
05:32:45  * SkyRocknRolljoined
11:22:43  * SkyRocknRollquit (Ping timeout: 248 seconds)
11:39:05  * SkyRocknRolljoined
13:04:04  * SkyRocknRollquit (Ping timeout: 240 seconds)
14:27:32  <rphillips>good morning
15:41:03  <creationix>mornin'
16:11:09  <creationix>hmm, lit server is down
16:11:35  <creationix>and my ssh connection just kicked me out. I wonder if the machine was rebooting
16:12:16  <creationix>no, 8 days uptime still, 0.06 load average
16:12:56  <creationix>must be a bug in the new network code I deployed yesterday
16:13:04  <creationix>restarting the lit process fixes it for now
16:13:38  <creationix>rphillips, how can I test a luvi process for resource leaks?
16:14:19  <rphillips>instruments is great on OSX
16:14:47  <creationix>true, I can hit it with lots of load locally and see if anything pops up
16:15:01  <creationix>I just know the production server on linux was getting 502 gateway errors from nginx
16:15:09  <creationix>the process was still running
16:20:03  <creationix>my guess is when I changed the logic for when handles close, something created a socket leak
16:22:22  <creationix>daurnimator, have you ever looked at ae (the redis event library) http://redis.io/topics/internals-rediseventlib
16:22:37  <creationix>it looks very simple and fast for times when libuv is too much
16:27:13  <creationix>rphillips, The lit server has lots of FIN_WAIT2 and CLOSE_WAIT sockets and /proc/$PIC/fd lists lots of open fds
16:27:22  <creationix>even plain http requests are causing them
16:27:40  <rphillips>probably a socket handle not being closed
16:30:53  <creationix>looks like mostly CLOSE_WAIT which means the server isn't closing it's half
16:33:42  <creationix>Weird, can't reproduce it locally. Must be a linux thing or some behavior triggered by sitting behind nginx
16:35:00  <creationix>can't reproduce on local linux box either
16:39:27  <creationix>yep, sitting behind nginx reproduces the behavior
16:59:07  <creationix>found it. The lit http server wasn't sending FIN back when the client sent FIN
16:59:38  <creationix>should probably update lit to use weblit someday to consoldidate code (it has it's own custom http server right now)
17:06:24  <creationix>rphillips: something is wrong with the new luvi I think. I've been noticing strange things the last few days
17:07:12  <creationix>I made a new luvi app called wscat (netcat for websockets) and it takes a really long time to boot up (over 1000 ms on arm)
17:07:19  <creationix>strace just shows a bunch of mmap
17:07:26  <rphillips>hmm. could be luajit
17:07:38  <creationix>now lit's server is hanging when trying to match versions
17:07:42  <creationix>so I can't publish to lit anymore
17:07:55  <creationix>(though I did fix the fd leak in http, that's unrelated)
17:08:22  <creationix>rphillips try to run wscat, install it with `lit make lit://creationix/wscat`
17:08:38  <creationix>and test it against lit with `wscat wss://lit.luvit.io/ lit`
17:12:36  * |--|quit (Ping timeout: 248 seconds)