What happens every 5 minutes?

[fabio998]

Hi Wandenberg and thanks for your module.

Every 5 minutes (min 0, 5, 15, 20, ....) the nginx-push-stream-module slowes down, the CPUs utilization for nginx processes goes up (from 5-10% to over 70-100%) and many publishers and subscribers get timeout error.

It seems like every 5 minutes the nginx-push-stream-module make something that cause a bottleneck.

We have about 1500 open channels.

Waiting for you, thanks Regards

[wandenberg]

Hi Fabio,

can you send me your configuration file? The module has some cron tasks, to know which one may be causing the problem I have to see in what mode are your configuration working.

Tell me what nginx and module version you are using too.

Regards, Wandenberg

[fabio998]

Hi Wandenberg,

thanks for your quick reply.

NGINX Version: 1.2.3 stable nginx-push-stream-module Version: the last one available on http://github.com/wandenberg/nginx-push-stream-module.git

Please note that the same server is also used as proxy. Also note that push_stream_subscriber_connection_ttl is disabled as an attempt to solve the issue.

This is my nginx.conf file: https://gist.github.com/3833400

Thanks again. Fabio

[wandenberg]

Hi Fabio,

are you sure that you do not have any cron task configured at your server? or any other module that could be causing the problem?

I checked the module again and we have 4 periodic tasks:

• ping message, is an individual timer, one to each request, and execute on the interval set on push_stream_ping_message_interval directive, in your case, each 10 seconds
• disconnect timer, is an individual timer, one to each request, and execute after the time configured on push_stream_subscriber_connection_ttl or push_stream_longpolling_connection_ttl directives, in your case, 30 seconds after the user connects
• buffer cleanup, is a timer responsible for collect expired messages from channels, and is executed each 5 seconds
• memory cleanup, is a timer responsible for clean the memory not being used anymore, and is executed on a timer based on push_stream_shared_memory_cleanup_objects_ttl directive, the default value (min value), which is your case, is each 4 seconds

So I don't think the culprit is the push stream module. Try to execute an strace on your worker to see what it is doing when the problem happens.

Regards, Wandenberg

[wandenberg]

Hi Fabio,

any news about that?

[fabio998]

Hi Wandenberg, sorry for the late response. I don't know what happened a few months ago, but I installed a new server with nginx and your module and all seems fine. So you can close this ticket. Thanks a lot for the support. Your module works really good with thousand of users. Best, Fabio