The web in a box - a next generation web framework for the Perl programming language

IRC log for #mojo, 2016-09-19

| Channels | #mojo index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:42 disputin joined #mojo
01:02 itaipu joined #mojo
01:20 ivi joined #mojo
01:29 Bert joined #mojo
01:32 itaipu joined #mojo
02:17 noganex_ joined #mojo
02:34 zivester joined #mojo
04:06 polettix joined #mojo
04:32 lluad joined #mojo
06:09 inokenty-w joined #mojo
06:11 dod joined #mojo
06:16 dod joined #mojo
06:22 dod joined #mojo
06:48 mbudde joined #mojo
07:06 AndrewIsh joined #mojo
07:21 Vandal joined #mojo
07:36 trone joined #mojo
08:12 abra joined #mojo
08:38 osfabibisi joined #mojo
08:52 rubin joined #mojo
09:03 kes joined #mojo
09:27 jlaffaye joined #mojo
09:28 jlaffaye I have a problem whith the systemd example unit file for hypnotoad, on reload, systemd does not see the new pid
09:28 jlaffaye I wonder if there is a race condition between hypnotoad for hot deploy and the new server writing its pid file
09:29 irqq joined #mojo
09:31 jlaffaye because if I read the pid file after the reload, the new pid is in there
09:52 kwa_ jlaffaye: I think I ran into that problem, too. Don't think I ever got to the bottom of it, so if you find a solution, ping it to me please.
10:06 jlaffaye kwa: well, one solution could be to wait for the old pid to exit, but AFAIK there is no easy way to do that on linux if you are not the parent of the process
10:07 dod joined #mojo
10:07 sri mojolicious version?
10:08 jlaffaye 6.13
10:09 alilles joined #mojo
10:11 jlaffaye but I just read the code of hypnotoad, it just send a signal to the old process, no synchronisation here
10:17 dotan_convos joined #mojo
10:19 sri that's your problem
10:19 sri the systemd unit files are for newer versions
10:19 kaare_ joined #mojo
10:20 sri you're over a year behind
10:21 jlaffaye yes, I use debian :/
10:22 sri i fixed systemd bugs when i added the examples to the docs
10:27 bobkare jlaffaye: what package is that? jessie has 5.54 and stretch 7.01
10:31 jlaffaye sri: I guess this is the fix: https://github.com/kraih/mojo/commit/25471294ef714e2116dd2bbe98de4893285c1121
10:33 jlaffaye bobkare: oh, right, its a custom package from $WORK
10:35 kwa Ah, good to know.
11:00 jamesaxl joined #mojo
11:06 itaipu joined #mojo
11:18 coolo joined #mojo
11:40 itaipu joined #mojo
11:40 dod joined #mojo
12:08 anton joined #mojo
12:16 trobotham joined #mojo
12:33 gizmomathboy joined #mojo
12:51 zivester joined #mojo
12:51 Vandal joined #mojo
12:53 CW joined #mojo
13:04 itaipu joined #mojo
13:09 sri ok, minion release day
13:09 sri looks like there are no more feature requests or complaints
13:11 sri maybe any opinions on minion worker state information? https://github.com/kraih/minion/issues/42
13:11 sri aside from max concurrent jobs and list of named queues, i imagine it could also include information about load on the worker
13:12 sri that could help with scaling
13:15 coolo batman: I adapted the test case - it should fail now everywhere
13:16 batman coolo: okey
13:18 jberger sri: How would you define load?
13:18 batman coolo: do you have any time trying to figure it out myself? i'm swamped :(
13:19 jberger batman / coolo : did i miss something? What are you talking about?
13:19 sri jberger: information on recent idle time
13:19 batman jberger: assetpack
13:19 coolo jberger: batman and me have a side project :)
13:20 sri jberger: not thought of an actual formula yet, but something that takes max concurrent jobs into account and how often that number has not been reached in the last few hours
13:20 jberger sri: so the idea is that if the worker hasn't had idle time in a while it might be time to consider more hardware
13:21 jberger Or with some formula
13:21 jberger But to the same end
13:21 sri yea, or increasing concurrency
13:21 jberger Will yes that first
13:21 jberger Well even
13:21 coolo batman: not really - but at least I know why travis succeeded: it skips sass
13:21 jberger sri: i could see that as being useful
13:22 batman coolo: ah! of course. should've remembered that
13:22 jberger Something to put into monitoring
13:22 coolo batman: I live fine with my work around. But #92 is still alive, but you can keep it closed and keep #105 alive
13:22 sri jberger: basically, i'm looking for more useful worker information to make the feature worth the cost
13:22 itaipu joined #mojo
13:23 jberger The cost is some persistence?
13:23 batman coolo: right. so #105 is not a deal breaker?
13:23 sri updating a json field with possibly lots of information every minute
13:23 sri together with the heartbeat
13:24 coolo batman: I redocument my workaround for people falling over the issue
13:24 batman nice
13:24 jberger sri: different idea ...
13:24 jberger Add a column on the worker table
13:24 jberger That gets updated with the heartbeat
13:24 sri umm
13:24 jberger Current load or so
13:25 sri that's already the idea
13:25 sri lol
13:25 jberger Ok then ;]
13:25 sri add a json column to the worker table, and update it with the heartbeat
13:25 sri at first i even considered merging host and pid information into it
13:26 sri but that's just wasteful
13:26 jberger Right but why does it even need to be json?
13:26 sri what else?
13:26 sri you want to be able to add more information when you need to, no?
13:27 jberger Surely just a number is more efficient, and you can add more columns later if you have to
13:27 jberger This seems like actual relational data
13:27 sri i don't want that many migrations
13:27 sri ok, i take that as a good argument against the feature :)
13:28 jberger Haha
13:29 sri in fact, i think i'll make an extremely well thought out proposal with all the different metrics we might want to report for workers a requirement for the feature
13:29 sri even though that means it will prolly never happen
13:32 nic When you measure the time that max wasn't hit, you're essentially looking at the time gaps between peak usage, which isn't hugely useful
13:33 nic Combining that with average number of busy workers makes it much more useful
13:33 nic (the principle from maths is you're then measuring the area under the graph)
13:35 jberger nic: it would have to be in combination with some monitoring
13:35 jberger Minion outputs the current values and graphite or whatever does the integration
13:36 Dandre joined #mojo
13:36 nic that makes sense
13:36 nic I think min, max, mean are all I'd want there
13:37 * jberger gets ready for work
13:37 jberger Biab
13:37 nic It does mean you can't query that from the commandline tho
13:37 sri right now you can only see what jobs a worker is curently processing
13:37 ramortegui joined #mojo
13:37 sri but you have no idea how overloaded a worker really is
13:38 sri the idea is to tell you that the worker has been at peak for the last few hours
13:38 sri and you can see what queues it is handling, and how many jobs are still unprocessed in the queue
13:38 sri and then figure out how much more hardware you need to add
13:38 nic understood
13:39 nic would you be able to see that measure of peak usage at the commandline?
13:39 sri hmm, i guess this goes more into the history data territory again https://github.com/kraih/minion/issues/26
13:40 sri nic: all information will always be available at the command line
13:40 sri an admin ui will only be an alternative to the command line interface
13:40 sri never a replacement
13:41 sri guess what *i* really want is history data, and not so much worker state data ;p
13:41 nic cool.  I see from those issues that everything I'd be wanting is already mentioned/considered
13:42 nic likewise
13:50 nic This might just be bonkers, but perhaps there could be a silo backend (postresql, influxdb, mysql, ...) with a common API
13:50 nic (actually, I'd best be quiet.  I've been away for a while and I'm sure to repeat stuff that's already been covered)
14:02 mcsnolte joined #mojo
14:15 sri anyway, there won't be any quick decisions with this
14:15 sri so, minion 6.0 release time \o/
14:19 sri Grinnz: *poke*
14:34 sri https://twitter.com/kraih/status/777875825005826048
14:55 zivester joined #mojo
14:59 ptolemarch joined #mojo
15:53 disputin joined #mojo
16:02 lluad joined #mojo
16:09 itaipu joined #mojo
16:28 disputin joined #mojo
17:10 PryMar56 joined #mojo
17:38 irqq joined #mojo
17:40 dod joined #mojo
17:57 go|dfish o
18:06 sri o
18:09 jberger I'm assuming you are both making faces at javascript as I realize that javascript doens't have s?printf :o
18:09 preaction it does now. it has template literals
18:22 sri O
18:22 jberger and for the record, that solves interpolation, not formatting
18:22 * jberger understands leftpad npm package now
18:32 dod joined #mojo
18:40 Grinnz hm... POST to pause timing out
18:42 Grinnz pause in general timing out :s
18:54 sri guess i'll wait with releasing mojolicious :o
18:55 sri jberger: actually, you want the feature, why don't you release it?
18:55 jberger 303?
18:56 jberger I can release it, that's no problem
18:56 sri s/feature/bug fix/
18:56 jberger been a while since I've done it
18:56 jberger but that's probably worth doing
18:56 jberger (after work though
18:56 jberger )
18:56 sri we've got enough version numbers again, feel free to mess up a few times :)
18:56 jberger haha, I'll try not to
18:57 * jberger fails 98 times
19:04 sri btw. still loving the hyperx headset
19:05 sri will definitely be using it to listen to background music while programming
19:06 vicash does anyone use the 'tel:' URL schema in mojolicious ?
19:06 vicash can it be used to make calls from a web browser that is running on a desktop ?
19:06 vicash i know that it works well on a modern smartphone
19:07 itaipu joined #mojo
19:07 laidback_01 joined #mojo
19:10 preaction vicash: it does whatever the browser says to do. so, if the desktop can make calls, and the browser knows what to do to do that, then yes
19:11 mishanti1 vicash: Depends on what has been installed on the client. If you have skype installed it works flawlessly.
19:11 lluad vicash: It's not mojolicious specific. Some VOIP services provide plugins to support it on the desktop, but it's not magicvally universal.
19:11 vicash thank you lluad, mishanti1 preaction ... i will research the existing methods ...
19:38 ptolemarch joined #mojo
19:48 ptolemarch joined #mojo
20:28 Grinnz oops, i never actually published https://github.com/Grinnz/Mojo-IOLoop-Subprocess-Sereal
20:51 stephan48 sri: what did you pay for it?
20:51 stephan48 how much*
20:52 sri the store i got it from price matched amazon
20:57 stephan48 the connector towards the usb soundcard is a normal 4-contact 3,5 plug?
20:58 stephan48 so theoretically you could also attach it to a smartphone*thinking loudly* and according to the amazon reviews this should work without problems
20:58 stephan48 now that is awesome
20:58 sri it does
20:59 stephan48 i got tired waiting for propper usb audio on android smartphones
21:00 stephan48 so this might be a good additional benefit then.
21:10 batman joined #mojo
21:47 polettix joined #mojo
22:05 orev joined #mojo
23:29 disputin joined #mojo
23:48 orev joined #mojo
23:57 gizmomathboy joined #mojo

| Channels | #mojo index | Today | | Search | Google Search | Plain-Text | summary