Tuesday, 5 June 2012< ^ >
arjan has set the subject to: Zotonic - the Erlang CMS
Room Configuration

[00:05:56] hc joins the room
[00:16:53] hc leaves the room
[01:11:25] Shawn Garbett leaves the room
[03:06:58] Shawn Garbett joins the room
[03:10:38] Shawn Garbett leaves the room
[08:05:09] arjan joins the room
[08:50:44] arjan leaves the room
[09:47:06] maas.maarten.zeeman joins the room
[10:07:00] Marc Worrell joins the room
[12:21:22] hc joins the room
[12:32:32] <hc> @400000004fcddfd530e6ab4c 12:30:35.820 [info] SMTP Dropping message, unknown host for recipient: <<"inping@xpmachine.com">>
[12:32:48] <hc> can you explain what this error message means? zotonic isn't trying to act as a relay, is it? :)
[12:39:53] <Marc Worrell> nope
[12:40:13] <Marc Worrell> it just drops any incoming e-mail when it doesn't know which site (or module) can handle it.
[12:41:25] <hc> ah, cool. i'm currently getting hundreds of these in my logfiles
[12:41:37] <hc> no big deal then
[12:41:44] <hc> i was afraid i had misconfigured something
[12:41:46] <Marc Worrell> then someone is trying to see if you relay, which you don't :)
[12:42:03] <Marc Worrell> we don't have any relaying module, so it is very safe.
[12:42:18] <hc> heh
[12:42:39] <Marc Worrell> there is one module with 'relay' in it, but that one only relays mail to users. Like 'username@yoursite.com' is relayed to the email address of that user.
[12:44:04] <hc> oh? that's nice.
[12:57:27] <hc> i'm currently looking at the webmachine code serving partial requests. it seems the resource_file_readonly passes a file as a {stream,}, when it could also pass it as a {file,}, in which case webmachine wouldn't have to read in the whole file
[12:57:45] <hc> but it seems webmachine reads into memory the whole range requested, so that wouldn't be a big win for huge files (several GB) either
[13:01:02] <hc> best fix is probably to have webmachine handle {stream,} responses "smarter", what do you think?
[13:08:24] <Marc Worrell> i think so to, which shouldn't be too hard, as we are using our own webmachine fork, so we are free to update/change what we want.
[13:19:27] <Marc Worrell> I was also thinking to start adding the sendfile functions, as introduced in R15.
[13:19:42] <Marc Worrell> Of course, we need to stay backward compatible with R14.
[14:53:04] <Marc Worrell> Funny, Twitter is back to server side rendering. Exactly as we have been doing since we started Zotonic :)
[14:59:07] arjan joins the room
[14:59:38] arjan leaves the room
[15:00:28] <hc> oh, that's interesting!
[15:00:38] <hc> i'm trying to find my way through that webmachine code - not exactly well documented ;)
[15:01:30] <hc> btw - speaking of webmachine: i find it rather annoying it opens a new logfile every hour. also, it'd be nice if the vhost was stored in the logfile
[15:02:26] <hc> what do you think?
[15:08:19] <Marc Worrell> i agree!
[15:08:29] <Marc Worrell> it is very irritating, we need better logging for that one.
[15:08:51] <Marc Worrell> maybe we can add some logging per site and then with roll-overs (or however that is called)
[15:09:17] <Marc Worrell> or a completely other logger, is there anyone looking in those log files anyway?
[15:16:19] hc is doing that occasionally, i also have a small python util that sorts the referers and google queries
[15:16:37] <hc> expects a standard apache logfile, so i use cat * | ./myutil.py to invoke it ATM.
[15:22:32] <Marc Worrell> I was always dreaming of a more-or-less realtime interface to this kind of information. Though log files are very good for post-mortem analysis (even essential). That is also why we are adding that extra request id for every request.
[16:49:52] <hc> [x] successfully added websocket code to webzmachine. now to add some tests and clean it up...
[16:52:50] <hc> (especially performance tests ;-)
[17:22:56] <maas.maarten.zeeman> @marc Yeah I look at those logs. It is currently the way to get some hourly stats out of 0.8.0 :-) Just count the lines.
[17:25:47] <maas.maarten.zeeman> But it misses the stuff that goes through a websocket.
[17:26:45] maas.maarten.zeeman leaves the room
[18:38:12] maas.maarten.zeeman joins the room
[18:57:10] arjan joins the room
[19:03:16] arjan leaves the room
[19:06:28] arjan joins the room
[19:17:24] arjan leaves the room
[19:30:19] <Marc Worrell> For the hourly stats we should use real stats. Maybe we can already get the statz app (more or less) integrated.
[19:31:25] <Marc Worrell> Then we can also count messages/data coming through all sockets
[19:32:28] <maas.maarten.zeeman> Yes. It is hard to see what the server is doing. Especially if you have a lot of push messages being sent.
[19:32:53] <maas.maarten.zeeman> You mentioned folsom some time ago. That looked pretty nice.
[19:33:25] <maas.maarten.zeeman> That would allow one to easily build application level metrics too.
[19:34:55] <maas.maarten.zeeman> statz is from elastic or not?
[19:36:49] <maas.maarten.zeeman> Recently we hit another erlang limit... not ports but ets tables.
[19:37:39] <maas.maarten.zeeman> A default of 1400 is a bit on the small side if you are using an ets table hungry app like inets httpc
[19:40:14] <maas.maarten.zeeman> We also need a better way to deal with the erl vm config stuf.
[19:53:53] <hc> the webzmachine/master branch doesn't seem to work with zotonic/master
[19:55:47] <hc> anyway, at https://github.com/hce/zotonic branch sendfile and https://github.com/hce/webzmachine branch sendfile i've committed my attempt to make sendfile work with zotonic (this also fixed the partial request memory exhaustion problem for big requests, so long as sendfile is available)
[19:56:25] <hc> i'll do more tests later
[19:56:29] <hc> need to go now
[19:56:31] hc leaves the room
[20:02:18] arjan joins the room
[20:12:50] Marc Worrell leaves the room
[20:21:11] Marc Worrell joins the room
[21:25:56] arjan leaves the room
[21:30:13] hc joins the room
[21:54:22] arjan joins the room
[22:12:18] simon.smithies joins the room
[23:10:47] simon.smithies leaves the room
[23:23:20] hc leaves the room
[23:33:58] arjan leaves the room
Powered by ejabberd Powered by Erlang Valid XHTML 1.0 Transitional Valid CSS!