[asterisk-bugs] [JIRA] (ASTERISK-29880) PJSIP queues reach limits

[Spilios Georgakopoulos (JIRA)]

     [ https://issues.asterisk.org/jira/browse/ASTERISK-29880?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Spilios Georgakopoulos updated ASTERISK-29880:
----------------------------------------------

    Description: 
Hello,

we have built a Voicemail platform for Telecom services in a Telecom Client based on Asterisk (18.6) and Opensips (3.1.0). Below you can find some details for architecture:
•	SIP signaling from Core Network
•	2 proxy opensips VMs for the incoming calls (load balancer to asterisk VMs) and outgoing calls 
•	3 asterisk VMs for handling the calls with built dialplans based on phpagi (asterisk conf files and PHP-AGI files) – We call phpagi scripts to call procedures from the DB or for some other processes.
•	The Data for user settings are stored in external DB (ORACLE in different VM)
•	The data for file storage (voicemails etc.) are stored in NFS

We have (and still are) faced serious problems in the service (in queues of PJSIP processor) , when we transfer more traffic from the core system (increase of active calls/Asterisk), without this traffic being so huge in order to justify the problems in the voicemail service. The resources of Asterisk have 12 CPU cores, 32GB RAM without reaching above the 50% when the problem occurs.
This causes delays in call processing and the problem is increasing until the service cannot support any call. During this problem, the active calls in Asterisk are increasing when the queues are reaching the limit. (we have not any issue until ~100 active calls/Asterisk).

Have you ever faced the same problems?
We need your support since it is a production issue and affects the service.

Br,
Spilios G.

  was:
Hello,

we have built a Voicemail platform for Telecom services based on Asterisk (18.6) and Opensips (3.1.0). Below you can find some details for architecture:
•	SIP signaling from Core Network
•	2 proxy opensips VMs for the incoming calls (load balancer to asterisk VMs) and outgoing calls 
•	3 asterisk VMs for handling the calls with built dialplans based on phpagi (asterisk conf files and PHP-AGI files) – We call phpagi scripts to call procedures from the DB or for some other processes.
•	The Data for user settings are stored in external DB (ORACLE in different VM)
•	The data for file storage (voicemails etc.) are stored in NFS

We have (and still are) faced serious problems in the service (in queues of PJSIP processor) , when we transfer more traffic from the core system (increase of active calls/Asterisk), without this traffic being so huge in order to justify the problems in the voicemail service. The resources of Asterisk have 12 CPU cores, 32GB RAM without reaching above the 50% when the problem occurs.
This causes delays in call processing and the problem is increasing until the service cannot support any call. During this problem, the active calls in Asterisk are increasing when the queues are reaching the limit. (we have not any issue until ~100 active calls/Asterisk).

Have you ever faced the same problems?

Br,
Spilios G.


> PJSIP queues reach limits
> -------------------------
>
>                 Key: ASTERISK-29880
>                 URL: https://issues.asterisk.org/jira/browse/ASTERISK-29880
>             Project: Asterisk
>          Issue Type: Bug
>      Security Level: None
>          Components: pjproject/pjsip
>    Affects Versions: 18.6.0
>         Environment: RedHat Linux
>            Reporter: Spilios Georgakopoulos
>            Severity: Major
>
> Hello,
> we have built a Voicemail platform for Telecom services in a Telecom Client based on Asterisk (18.6) and Opensips (3.1.0). Below you can find some details for architecture:
> •	SIP signaling from Core Network
> •	2 proxy opensips VMs for the incoming calls (load balancer to asterisk VMs) and outgoing calls 
> •	3 asterisk VMs for handling the calls with built dialplans based on phpagi (asterisk conf files and PHP-AGI files) – We call phpagi scripts to call procedures from the DB or for some other processes.
> •	The Data for user settings are stored in external DB (ORACLE in different VM)
> •	The data for file storage (voicemails etc.) are stored in NFS
> We have (and still are) faced serious problems in the service (in queues of PJSIP processor) , when we transfer more traffic from the core system (increase of active calls/Asterisk), without this traffic being so huge in order to justify the problems in the voicemail service. The resources of Asterisk have 12 CPU cores, 32GB RAM without reaching above the 50% when the problem occurs.
> This causes delays in call processing and the problem is increasing until the service cannot support any call. During this problem, the active calls in Asterisk are increasing when the queues are reaching the limit. (we have not any issue until ~100 active calls/Asterisk).
> Have you ever faced the same problems?
> We need your support since it is a production issue and affects the service.
> Br,
> Spilios G.



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