| 01:21.14 | *** join/#asterisk-dev coreyfarrell (~coreyfarr@68-114-57-249.dhcp.davl.vt.charter.com) |
| 01:21.14 | *** mode/#asterisk-dev [+o coreyfarrell] by ChanServ |
| 04:38.41 | *** join/#asterisk-dev jkroon (~jkroon@41.114.17.235) |
| 04:53.48 | *** join/#asterisk-dev jkroon (~jkroon@41.114.17.235) |
| 04:57.19 | *** join/#asterisk-dev jkroon (~jkroon@41.114.17.235) |
| 05:40.47 | *** join/#asterisk-dev jkroon (~jkroon@41.169.49.18) |
| 06:18.30 | *** join/#asterisk-dev pchero_work (~pchero@87.213.247.82) |
| 06:22.52 | *** join/#asterisk-dev CELYA_ (~Thunderbi@2a01:e0a:d:1111:b749:768d:ed77:7c9a) |
| 06:30.11 | *** join/#asterisk-dev pchero_work (~pchero@87.213.247.82) |
| 06:48.24 | *** join/#asterisk-dev jkroon (~jkroon@41.169.49.18) |
| 07:15.39 | *** join/#asterisk-dev jkroon (~jkroon@41.169.49.18) |
| 07:34.18 | *** join/#asterisk-dev jkroon (~jkroon@41.169.49.18) |
| 07:45.03 | *** join/#asterisk-dev hehol (~hehol@gatekeeper.loca.net) |
| 07:53.42 | *** join/#asterisk-dev jkroon (~jkroon@41.169.49.18) |
| 08:01.47 | *** join/#asterisk-dev hehol (~hehol@gatekeeper.loca.net) |
| 08:24.19 | *** join/#asterisk-dev pchero_work (~pchero@87.213.247.82) |
| 09:01.50 | *** join/#asterisk-dev wraythe (~wraythe@li702-54.members.linode.com) |
| 11:44.59 | *** join/#asterisk-dev pc-m (~pcm@modemcable094.94-70-69.static.videotron.ca) |
| 12:39.56 | seanbright | 7500 mailboxes... cool |
| 12:42.28 | file | seanbright: just a few |
| 12:44.44 | seanbright | there are two options: fix the performance or remove the feature |
| 12:44.57 | seanbright | i sha'nt be adding a configuration option |
| 12:46.15 | file | orly |
| 12:46.28 | gtjoseph | I feel a -2 coming on. |
| 12:47.01 | seanbright | heh |
| 12:47.23 | file | wonders if their voicemail is on a share |
| 12:48.23 | gtjoseph | SAN :) |
| 12:48.48 | file | gtjoseph: we don't speak of that... |
| 12:53.11 | seanbright | do we know? |
| 12:53.38 | seanbright | do we know it's a SAN or was gtjoseph kidding? |
| 12:53.41 | gtjoseph | no but it's likely with that many mailboixes |
| 12:53.54 | gtjoseph | well, SAN or some kind of share |
| 12:56.54 | seanbright | wonderful |
| 13:00.06 | seanbright | probably the best fix would be to create some kind of lock index |
| 13:00.41 | seanbright | but i will need to think about that some |
| 13:01.37 | seanbright | i mean, if it takes 30 minutes to traverse 7500 directories, i think it would have to be on network attached storage |
| 13:07.26 | seanbright | i just saw this part: debian9 realtime odbc mysql |
| 13:07.45 | seanbright | hmmm |
| 13:14.28 | *** join/#asterisk-dev sduthil (~sduthil@ipagstaticip-cf899324-841a-2614-948a-7dbf21746b29.sdsl.bell.ca) |
| 13:42.18 | *** join/#asterisk-dev bford (uid283514@gateway/web/irccloud.com/x-fmpmafdrlljiuaag) |
| 13:42.18 | *** mode/#asterisk-dev [+o bford] by ChanServ |
| 13:53.02 | seanbright | i created 10000 VM directories and it took less than a second |
| 13:53.15 | seanbright | so i assume he is using network attached storage. |
| 13:53.20 | seanbright | background thread? |
| 13:55.10 | file | that seems... problematic... as if it's an async background process someone could then go into voicemail and you'd then potentially remove their lock |
| 13:55.19 | seanbright | great point |
| 13:55.25 | seanbright | so what is slow about the code? lstat()? |
| 13:55.32 | seanbright | that's my best guess |
| 13:56.26 | coreyfarrell | even on NAS 30 minutes seems excessive for 7500 dirs even on nas, that's 240ms each. |
| 13:57.32 | *** join/#asterisk-dev pchero_work (~pchero@51.247.195.35.bc.googleusercontent.com) |
| 13:57.53 | seanbright | so on linux i can skip the the lstat() call because the dirent will have the file type |
| 13:58.09 | seanbright | which means i would only need the full filename if it was a directory |
| 14:02.00 | *** join/#asterisk-dev jkroon (~jkroon@41.169.49.18) |
| 14:14.53 | *** join/#asterisk-dev jkroon (~jkroon@41.169.49.18) |
| 14:18.14 | seanbright | i have my VM spool directory on a tape drive, and it's a little slow |
| 14:18.43 | seanbright | each VM is stored in a punch card that we feed in manually |
| 14:29.18 | file | seanbright: ooh scalable |
| 14:29.47 | file | seanbright: yup - NFS |
| 14:29.52 | seanbright | fuck |
| 14:30.08 | file | better than Samba |
| 14:35.32 | *** join/#asterisk-dev kharwell (uid358942@gateway/web/irccloud.com/x-aqgtqflzlajrgkcb) |
| 14:35.32 | *** mode/#asterisk-dev [+o kharwell] by ChanServ |
| 14:47.14 | *** join/#asterisk-dev pchero_work (~pchero@51.247.195.35.bc.googleusercontent.com) |
| 14:57.27 | *** join/#asterisk-dev gtjoseph (~gtjoseph@unaffiliated/gtj) |
| 14:57.27 | *** mode/#asterisk-dev [+o gtjoseph] by ChanServ |
| 15:01.25 | coreyfarrell | looks like asterisk-bugs isn't getting mail from JIRA (unless it's just delayed). gtjoseph I responded to you on ASTERISK-28568 |
| 15:01.48 | gtjoseph | ack |
| 15:05.16 | gtjoseph | Ha! yeah 18mb, sorry |
| 15:08.32 | coreyfarrell | still it's RSS so I don't think it matters. freeing memory can increase RSS if it runs an object destruction function for the first time. |
| 15:12.11 | gtjoseph | yeah it's be nice to see the virtual |
| 15:12.18 | gtjoseph | yeah it'd be nice to see the virtual |
| 16:15.03 | *** join/#asterisk-dev tmoore (~tmoore@hades1-43.kettering.edu) |
| 16:32.01 | *** join/#asterisk-dev CELYA_ (~Thunderbi@2a01cb0583d31d00f4d7ddd1bc0987bc.ipv6.abo.wanadoo.fr) |
| 16:52.45 | *** join/#asterisk-dev CELYA_1 (~Thunderbi@lfbn-1-9707-146.w86-236.abo.wanadoo.fr) |
| 16:55.22 | *** join/#asterisk-dev csguth (c8af3d51@200.175.61.81.static.gvt.net.br) |
| 16:55.39 | csguth | Hi all |
| 16:56.20 | csguth | I'm developing an Asterisk 13 Channel Driver, but I'm not sure when should Asterisk call my driver's read function. |
| 16:56.44 | csguth | Do I have to create a FD and write on that? |
| 16:57.45 | file | if a file descriptor is placed on the channel then read will be called when there is data to be read on it |
| 17:00.16 | csguth | Hm, I see |
| 17:00.41 | file | otherwise the channel driver can use ast_queue_frame to queue frames |
| 17:01.03 | csguth | I'm my case, I'm using an external library to produce PCM audio data. |
| 17:01.57 | csguth | The lib produces PCM returns the audio data when the user (asterisk) calls a callback. |
| 17:01.58 | *** join/#asterisk-dev pc-m (~pcm@modemcable094.94-70-69.static.videotron.ca) |
| 17:02.42 | csguth | I'd like to have asterisk calling such a callback in the correct frequency (e.g., each 20ms). |
| 17:03.30 | csguth | One possible solution would be to create a timer, pass its FD to the channel and make it call the callback from the read function. |
| 17:03.49 | csguth | Another solution would be to create a timer to call ast_queue_frame |
| 17:03.57 | csguth | Idk which one is more appropriate |
| 17:04.14 | file | read is probably easier and more efficient |
| 17:04.32 | *** join/#asterisk-dev sduthil (~sduthil@ipagstaticip-cf899324-841a-2614-948a-7dbf21746b29.sdsl.bell.ca) |
| 17:04.36 | csguth | So does the timer thing make sense for you? |
| 17:05.22 | file | yes |
| 17:05.32 | *** join/#asterisk-dev sahmed (~sahmed@cpe-70-114-236-63.austin.res.rr.com) |
| 17:05.44 | csguth | Thanks!! |
| 17:06.05 | csguth | Thus the channel driver must control the frequency of the reads |
| 17:24.51 | seanbright | sigh |
| 17:27.53 | file | seanbright: yeah.... |
| 17:30.30 | coreyfarrell | isn't odbc storage the only (theoretically) safe way to have voicemail storage shared between multiple asterisk servers? |
| 17:31.07 | coreyfarrell | well I guess imap too... but not file storage |
| 17:32.37 | *** join/#asterisk-dev tmoore (~tmoore@hades1-43.kettering.edu) |
| 17:34.17 | file | does not know |
| 17:34.25 | file | I question any of it as being safe |
| 17:43.14 | seanbright | the only correct 'fix' is to find a better way to lock |
| 17:43.38 | seanbright | 'lock' directory at the root of VM_SPOOL_DIR that contains the lock files |
| 17:43.51 | seanbright | the lock files will have the pid of the asterisk process that locked them |
| 17:43.59 | seanbright | nah, even that wouldn't work |
| 17:44.00 | seanbright | fuck it |
| 17:44.06 | seanbright | configuration for the win |
| 17:48.25 | *** join/#asterisk-dev jkroon (~jkroon@165.16.204.106) |
| 17:53.38 | *** join/#asterisk-dev Stefan26 (~Stefan26@78-67-218-242-no207.tbcn.telia.com) |
| 18:00.13 | csguth | file, the timer + read worked like a charm |
| 18:00.18 | csguth | thank you very much |
| 18:00.25 | file | aye |
| 18:22.35 | seanbright | i'm being way too knee-jerky on this voicemail thing |
| 18:26.18 | seanbright | if i'm being honest, i would much rather just remove the functionality than to put it behind a configuration option |
| 18:26.54 | Stefan26 | seanbright, you worked with asterisk development for fun only? |
| 18:27.00 | gtjoseph | why was it added in the first place? I didn't get a chance to look at the original issue. |
| 18:27.07 | seanbright | i don't know that it adds much value when `find /var/spool/asterisk/voicemail -name '.lock' -delete` would do the job |
| 18:27.24 | seanbright | gtjoseph: it was an open issue and i was trying to close open issues :) |
| 18:27.36 | seanbright | > We recently experienced a segfault while a user was accessing their voice mail messages. This resulted in Asterisk leaving behind a .lock file in the user's voice mail directories. This prevented them from accessing their voice mail or receiving new voice mails until we noticed it. Can Asterisk automatically delete these lock files when it starts up? |
| 18:28.02 | gtjoseph | ah. |
| 18:28.24 | seanbright | Stefan26: it's fun, yes. but i also use asterisk at work. |
| 18:28.42 | gtjoseph | well, I think you could reasonably have folks add the find -delete to an ExecStartPre in a service file. |
| 18:28.50 | gtjoseph | if they needed it. |
| 18:29.45 | gtjoseph | with appropriate change/upgrade notes. |
| 18:29.47 | Stefan26 | I used to work for a small telecom company that used asterisk in their telephony solutions. I enjoyed browsing asterisk-code (first practical instance of C-programming for me) |
| 18:30.17 | coreyfarrell | yeah having asterisk itself never delete the voicemail .lock files seems like a good option. |
| 18:30.40 | seanbright | gtjoseph: yeah. i'm going to think on it for a bit, but that is the direction i am leaning. |
| 18:31.03 | gtjoseph | i'd be fine with it. |
| 18:31.07 | seanbright | is the idea behind lock files that you might have more than 1 asterisk instance writing to the same place? |
| 18:31.28 | gtjoseph | or some external thing that can manipulate mailboxes |
| 18:31.28 | seanbright | if not, why couldn't the locks just be in-memory ones? |
| 18:31.33 | seanbright | gotcha |
| 18:31.44 | gtjoseph | kinda goes with polling. |
| 18:31.57 | seanbright | ah yes. |
| 18:32.07 | Stefan26 | "we" built our own voicemail solution without app_voicemail... we didn't trust it |
| 18:32.18 | Stefan26 | so we used some combo of dialplan and Record |
| 18:32.29 | seanbright | we do something similar here |
| 18:32.31 | Stefan26 | (on asterisk 13) |
| 18:33.03 | seanbright | not because we didn't trust it, just because all of our business logic is outside of asterisk, so it didn't make a lot of sense |
| 18:33.34 | Stefan26 | there's no way i can use asterisk as a part of my new company's solutions... they have weird policies |
| 18:34.32 | Stefan26 | but that might mean it's time to start a hobby project, writing a silly module |
| 18:35.14 | seanbright | so assuming we roll out the code that finds the lock files, do we re-open ASTERISK-20207? |
| 18:50.28 | gtjoseph | seanbright: I don't think we need to re-open it but a comment on it that we're removing it in favor of instructions to DIY would be OK. |
| 18:59.43 | *** join/#asterisk-dev sduthil (~sduthil@ipagstaticip-cf899324-841a-2614-948a-7dbf21746b29.sdsl.bell.ca) |
| 19:04.42 | seanbright | gtjoseph: so the inclusion of the feature was never in CHANGES, would you still like a note about its removal? |
| 19:05.21 | gtjoseph | yes please. let's cover our asses, we bases. |
| 19:05.32 | gtjoseph | s/we/our/ |
| 19:05.52 | gtjoseph | no that;s not what I meant but you get the idea |
| 19:23.49 | seanbright | gtjoseph: cherry-pick to 17.0? |
| 19:24.16 | gtjoseph | no. i think we're too close for that. |
| 19:24.21 | seanbright | grr |
| 19:24.21 | seanbright | ok |
| 19:24.33 | gtjoseph | unless file wants to chime in |
| 19:24.50 | file | I'd err on not doing so |
| 19:25.09 | file | it'll land 17.1 |
| 19:25.37 | Stefan26 | did asterisk 16 or 17 include any webrtc improvements? |
| 19:25.51 | seanbright | dozens |
| 19:29.21 | coreyfarrell | seems weird if 17.0.0 will contain the delete .lock "feature" and we know it's getting removed from 17.1.0. |
| 19:30.47 | *** join/#asterisk-dev AsteriskRoss (~AsteriskR@r01.nt-r1.nor.gb.voicehost.co.uk) |
| 19:31.21 | seanbright | i didn't know what else to call it |
| 19:31.28 | seanbright | "option?" |
| 19:31.56 | seanbright | i tend to agree though. it's a single function and one call to it. |
| 19:32.12 | file | it should technically trigger an rc3 |
| 19:32.18 | file | if it goes in |
| 19:32.41 | file | it's also not going to land in the latest releases either |
| 19:33.02 | seanbright | i mean... ok |
| 19:33.05 | file | (unless we pulled it into them too) |
| 19:33.09 | seanbright | whatever you guys have to do |
| 19:33.57 | file | what do others think? |
| 19:34.21 | coreyfarrell | my vote as someone who does not have to deal with generating releases is that deleting .lock files on startup is a regression. |
| 19:34.29 | seanbright | hmmm\ |
| 19:34.34 | seanbright | i don't agree with th at |
| 19:34.34 | coreyfarrell | if you have two instances of asterisk sharing the same VM storage they should respect each others lock's, otherwise an in-memory lock would be good enough. |
| 19:34.44 | file | how long was the change in? |
| 19:35.05 | seanbright | apr 12 |
| 19:36.14 | coreyfarrell | ok if it's been in multiple releases maybe just treat it as a bug. |
| 19:36.53 | file | so it's been in since 16.4.0 |
| 19:37.07 | coreyfarrell | sorry not sure where I got the impression that it was new to the next releases |
| 19:37.39 | seanbright | ok, 17.1 it is |
| 19:39.09 | file | next thing up on the docket - an official Asterisk Docker image |
| 19:39.12 | file | j/k |
| 19:56.55 | Stefan26 | I built my own rpm for asterisk, had no idea what i was doing |
| 19:56.58 | Stefan26 | but it worked |
| 22:23.06 | *** join/#asterisk-dev spatel (~spatel@pool-98-118-124-216.bstnma.fios.verizon.net) |