[asterisk-bugs] [JIRA] (ASTERISK-30401) res_pjsip: Addition of endpoint/AOR requires notification for qualify/contact status

[Joshua C. Colp (JIRA)]

    [ https://issues.asterisk.org/jira/browse/ASTERISK-30401?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=261226#comment-261226 ] 

Joshua C. Colp commented on ASTERISK-30401:
-------------------------------------------

Actually it looks like I didn't add an AMI action, just a CLI command that can be executed via AMI. There's "pjsip reload qualify endpoint" and "pjsip reload qualify aor".

> res_pjsip: Addition of endpoint/AOR requires notification for qualify/contact status
> ------------------------------------------------------------------------------------
>
>                 Key: ASTERISK-30401
>                 URL: https://issues.asterisk.org/jira/browse/ASTERISK-30401
>             Project: Asterisk
>          Issue Type: Bug
>      Security Level: None
>          Components: Resources/res_pjsip
>    Affects Versions: GIT, 18.9.0
>         Environment: Prod: RHEL 8 (some Kubernetes thing that I couldn't possibly describe)
> Test: Ubuntu 22.04 on Intel NUC Skull Canyon with 16GB of RAM and a simple SSD
>            Reporter: Jonathan Rose
>            Severity: Major
>         Attachments: extconfig.conf, mysql output.txt, pjsip.conf, res_odbc.conf, sorcery.conf
>
>
> Realtime PJSIP endpoints/aors have a bit of a crippling problem right now... if I have a pair of endpoint/aor present at the point when res_pjsip.so is loaded and register afterwards, registration works fine and device state is set to 'Not In Use' as you would expect it to.
> If, however, I add a new pair of endpoint/aor to the realtime backend and register *that* user, the registration will appear to work correctly from a SIP perspective, but within Asterisk the device state will be locked to unavailable. I can unregister and reregister any number of times at that point and the device state will remain unavailable.
> Issuing 'module reload res_sip.so' from the CLI will 'fix' the issue, but this is realtime... that's not really an acceptable workaround. I'd argue that it violates the whole concept of Realtime and I'm kinda shocked I couldn't find any reports of this happening already.
> I've confirmed that this problem seems to be backend agnostic, having used both odbc and curl as the realtime backends.
> I will provide minimalist configuration for reproducing the issue.
> Confirmed within 18.9.0-cert1 as well as the current (as of 1/20/2023) master branch of github.



--
This message was sent by Atlassian JIRA
(v6.2#6252)