[asterisk-dev] [Code Review] 3362: func_periodic_hook: New function for periodic hooks.
Corey Farrell
reviewboard at asterisk.org
Sat Mar 29 19:12:00 CDT 2014
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviewboard.asterisk.org/r/3362/#review11441
-----------------------------------------------------------
Do you have any ideas on how this will effect CPU load?
I was thinking of this as an example:
A system with 1 inbound call per second, each call lasting 5 minutes. Every inbound channel gets a 60 second timer for beep Playback (assume the timer runs 5 times). The inbound channel then dials outbound. You would reach about 600 concurrent "real" channels (300 in, 300 out), 2 real channels created and destroyed every second.
Of the 300 timers on inbound channels, 5 will fire every second. This means 10 local channels are created and destroyed every second. This gives us a total of 610 concurrent channels, 12 channels created and destroyed every second. I know channel creation/destruction isn't a lightweight operation, but I don't know if it will have a big impact overall.
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21128>
Need to document and enforce a minimum interval. Also need to document that interval is whole seconds only.
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21114>
Can we get a documented variable for the original channel name or uniqueid? I'd prefer something defined by this module (so not SPY_CHANNEL).
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21127>
state->interval could be added to hook_thread_arg so we can use it for dial timeout.
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21109>
state->last_hook is never pre-initialized to ast_tvnow(), so hooks would always trigger on first check after being created or re-enabled. Unless this behaviour is documented I would expect the first run after create or resume to be delayed state->interval seconds.
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21121>
If do_hook fails we should log an error but still update state->last_hook. Otherwise we would try running do_hook for every audio frame until it succeeds.
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21120>
This confused me for a minute, hook_on is run when the hook is created, hook_re_enable is called when someone writes "on".
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21126>
Need checks for !ast_strlen_zero on all args. Maybe also a verify the exten at context exists?
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21111>
Can we detach the audio hook here? I know we escape the hook callback very early but it still feels wasteful to keep it enabled.
Possibly remove state->disabled variable?
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21125>
Check for NULL chan here.
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21122>
Check for NULL chan here too
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21112>
Should we have an option to delete a hook? Something to free all memory from a periodic_hook that will not be used again.
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21118>
AST_MODULE for registrar?
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21119>
AST_MODULE for registrar?
/trunk/funcs/func_periodic_hook.c
<https://reviewboard.asterisk.org/r/3362/#comment21116>
Is this due to a technical limitation of ChanSpy? If not we should incorporate hook_id into the group variable so we limit 1 channel per hook_id.
Maybe add to documentation that a timer interval will be skipped if the channel from the previous interval hasn't ended.
- Corey Farrell
On March 29, 2014, 4:27 p.m., Russell Bryant wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviewboard.asterisk.org/r/3362/
> -----------------------------------------------------------
>
> (Updated March 29, 2014, 4:27 p.m.)
>
>
> Review request for Asterisk Developers.
>
>
> Repository: Asterisk
>
>
> Description
> -------
>
> This commit introduces a new dialplan function, PERIODIC_HOOK().
> It allows you run to a dialplan hook on a channel periodically. The
> original use case that inspired this was the ability to play a beep
> periodically into a call being recorded. The implementation is much
> more generic though and could be used for many other things.
>
> The implementation makes heavy use of existing Asterisk components.
> It uses a combination of Local channels and ChanSpy() to run some
> custom dialplan and inject any audio it generates into an active call.
>
> The other important bit of the implementation is how it figures out
> when to trigger the beep playback. This implementation uses the
> audiohook API, even though it's not actually touching the audio in any
> way. It's a convenient way to get a callback and check if it's time
> to kick off another beep. It would be nice if this was timer event
> based instead of polling based, but unfortunately I don't see a way to
> do it that won't interfere with other things.
>
>
> Diffs
> -----
>
> /trunk/funcs/func_periodic_hook.c PRE-CREATION
> /trunk/CHANGES 411572
>
> Diff: https://reviewboard.asterisk.org/r/3362/diff/
>
>
> Testing
> -------
>
> Called the following extension (100 at test), both letting it run all the way through, as well as hanging up at various points in the middle.
>
> [hooks]
>
> exten => beep,1,Answer()
> same => n,Playback(beep)
>
> [test]
>
> exten => 100,1,Answer()
> same => n,Set(BEEP_ID=${PERIODIC_HOOK(hooks,beep,5)})
> same => n,Wait(20)
> same => n,Set(PERIODIC_HOOK(${BEEP_ID})=off)
> same => n,Wait(20)
> same => n,Set(PERIODIC_HOOK(${BEEP_ID})=on)
> same => n,Wait(20)
> same => n,Hangup()
>
>
> Thanks,
>
> Russell Bryant
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20140330/c880ea03/attachment-0001.html>
More information about the asterisk-dev
mailing list