Hi Cindy,

 

There are three ingress controllers. One for kube-system services, one for openstack services, one for mariadb dedicatedly.  With current simplex deployment, by default, there are no ingress created in kube-system namespace (I didn’t see it in my deployment). In openstack namespace, I can see many ingresses created, like aodh, barbican, ceilometer, cinder, glace, horizon, nova, keystone, neuron, etc.

 

Regarding frequency, I am unsure. That’s why I initialize this discussion, and hope to get feedbacks from community experts.

 

Thanks.

Yi

From: Xie, Cindy
Sent: Monday, May 20, 2019 10:12 PM
To: Wang, Yi C <yi.c.wang@intel.com>; Rowsell, Brent <Brent.Rowsell@windriver.com>
Cc: starlingx-discuss@lists.starlingx.io
Subject: RE: [Starlingx-discuss] About Nginx worker process number

 

Yi,

What service request will Nginx accept and process, and on what kind of frequency? Just need to evaluate how it is being used and the required responsiveness.

 

Thx. - cindy

 

From: Wang, Yi C [mailto:yi.c.wang@intel.com]
Sent: Saturday, May 18, 2019 10:54 PM
To: Rowsell, Brent <Brent.Rowsell@windriver.com>
Cc: starlingx-discuss@lists.starlingx.io
Subject: Re: [Starlingx-discuss] About Nginx worker process number

 

Hi all,

 

Any recommends on the setting for the number of  nginx worker processes of ingress controllers? Thanks.

 

BR.

Yi

From: Wang, Yi C [mailto:yi.c.wang@intel.com]
Sent: Thursday, May 16, 2019 8:50 PM
To: Rowsell, Brent <Brent.Rowsell@windriver.com>
Cc: starlingx-discuss@lists.starlingx.io
Subject: [Starlingx-discuss] About Nginx worker process number

 

Hi Brent,

 

I am working on the bug https://bugs.launchpad.net/starlingx/+bug/1823803 which is about large nginx worker process number. With current starlingx code, there are three nginx ingress controllers, one for kube-system namespace ingress, one for openstack namespace ingress, and the last one for mariadb.  Each controller creates  as many worker processes as the number of cores.

 

I have initialized three patches to fix this issue. I have no idea about how much the workload of the three ingress controllers will be, so I temporally set the number of worker processes as 1.  Do your experts have any suggestions on this setting? Thanks.

https://review.opendev.org/#/c/659456

https://review.opendev.org/#/c/659533

https://review.opendev.org/#/c/659464

 

BR

Yi