Intel® Collaboration Suite for WebRTC
Community support and discussions on the Intel® Collaboration Suite for WebRTC (Intel® CS for WebRTC).

Clustering Strategies MCU 4.2.1!!!!!!

Naresh_R_1
New Contributor I
1,603 Views

Dear Team,

We have configured MCU 4.2.1 on CentOS Linux release 7.6.1810 (Core) 

Server 1: RabbitMq, MongoDB and Cluster-Manager — Private Network 

Here we are using default cluster_manager strategies 

 

Server 2: Management API , Management console and Basic APP Configured - Public

RabbitMQ and MongoDB host set to server 1 IP Address 

 

Server 3: Portal and WebRTC Agent  Configured  - Public 

portal.toml and webrtc-agent/agent.toml pubic IP is provided 

And RabbitMQ and MongoDB host set to server 1 IP

 

Server 4: Audio/Video/WebRTC/Conference-Agent  Configured- Public 

webrtc-agent/agent.toml pubic IP is provided 

And RabbitMQ host set to server 1 IP

 

Now when testing basic app I can see whole load is moving to Server3 WebRTC agent only 

 

My question here is how we distribute load to server 4 WebRTC agent. Tried all cluster strategies but could not work 

How can we achieve load balancing here?

 

When more users are logged in only server 3 WebRTC agent which is last used is getting utilised and CPU is raking to 80%

I want to distribute this load to different WebRTC agents how can we achieve this?

 

How can we move to another WebRTC agent server when one of the webRTC agent load is high. Since we need to put WebRTC agent in auto scaling mode in AWS when load is high we can spin up another agent automatically so that load is getting distributed 

 

Also /?forward=true is only working, when we try to open publish with mix we are not able to see farend videos in mix stream but able to publish to forward stream users.

 

Thanks

Naresh

 

 

0 Kudos
13 Replies
Naresh_R_1
New Contributor I
1,603 Views

Dear Team,

 

Any update on this?

Thanks

Naresh

0 Kudos
Li_C_Intel
Employee
1,603 Views

cluster_manager.toml:

Choose “least-used” or “round-robin” or “randomly-pick” instead of the default “last-used” for webrtc

0 Kudos
Naresh_R_1
New Contributor I
1,603 Views

Dear Team,

I have 2 servers in each audio/video agents installed  each with cluster solution and both max load is set to max_load = 0.20 #default: 0.85Now when i am using basic example application only with one single room only 1 server is getting utilised and that too it crossed the max_load. as you told  he worker exceeds the max load, then it will not be treated as available worker.

here then load should go to next available worker right? please explain how this will workl

 

Thanks

Naresh

0 Kudos
Naresh_R_1
New Contributor I
1,603 Views

From the above only one worker is getting utilised when we login to single room even it exceeded the max_load. 

Cluster Strategy is set as below

audio= most_used

video=least_used

As per the configuration when any when 1 worker exceeds the max_load it should automatically switch to another available worker, in this case it is not working as expected.

When new user joins to new room then load is shifted to another worker.

Please let me know how we can achieve this even in single room there are so many users and work_load sis exceeded it should shift to another worker

Thanks

Naresh

0 Kudos
Naresh_R_1
New Contributor I
1,603 Views

Dear Team,

Any update on the above mentioned query will be appreciated.

Thanks

Naresh

0 Kudos
Li_C_Intel
Employee
1,603 Views

Hi,

One room's single audio/video mixer will always run on same process.

0 Kudos
N__Aji
Novice
1,554 Views

One room's single audio/video mixer will always run on same process. : - Is there any way to scale this process or worker ?

Any update would be appreciated.

Thank you.

0 Kudos
Naresh_R_1
New Contributor I
1,603 Views

Hi,

My question here is how can we handle if there are so many users like 100 and above in a single room  then 1 single worker cannot handle that much cpu resources. how we can achieve load balancing with single room with many users

Is it the same with forward room also?

 

Thanks

Naresh

0 Kudos
Naresh_R_1
New Contributor I
1,603 Views

Any Update on the above query please please

 

Thanks

Naresh

0 Kudos
Chandramouli_P
Beginner
1,603 Views

Hello Team,

Any update would be appreciated.

Thank you.

Best Regards,
Chandramouli.

0 Kudos
Chandramouli_P
Beginner
1,603 Views

Hello Team,

Any help would be appreciated.

Thank you.

Best Regards,
Chandramouli.

0 Kudos
Li_C_Intel
Employee
1,603 Views

Hi,

The video node's cpu is mainly used by encoding and decoding. So the increase of users has little affect on audio/video nodes' cpu usage if they only subscribe.

The forward room is different.

0 Kudos
N__Aji
Novice
1,603 Views

We have used MCU 4.3 version.

We are also facing load balancing issue in audio video agents mentioned above. Even if we use multiple audio/video agents after maximum load utilisation, it is not connecting to other agents. Is there any way to point a particular room to a particular A-V agent ? So that if we have two A-V agents, then launching 1st room will always connect to particular A-V agent and 2nd room to another A-V agent.

Please advise whether we have any way to do it.

Any help would be appreciated.

Thanks

 

0 Kudos
Reply