On Mon, Apr 29, 2019 at 10:29 PM Rowsell, Brent <Brent.Rowsell@windriver.com> wrote:
Victor,
See inline
Brent
-----Original Message----- From: Victor Rodriguez [mailto:vm.rod25@gmail.com] Sent: Monday, April 29, 2019 9:43 AM To: Miller, Frank <Frank.Miller@windriver.com> Cc: starlingx-discuss@lists.starlingx.io Subject: Re: [Starlingx-discuss] Agenda: Weekly Containerization Meeting Monday April 29
Hi Frank
I am in charge of the track of performance and footprint of the STX systems
Due to the fact that this bug was a footprint issue I was wondering if should discuss on the meeting on what stages and part of the STX do you want me to track the memory footprint
Also, I have a question regarding the chance to set up -m to each docker image so we can limit the amount of memory of each one. One of the experiments we did las Friday debugging the issue was that if we set -m and the system has SWAP it will take memory from there and keep running since we in STX do not have swap the containers fails from starvation. [BR] Enabling SWAP, in general, is not an option for performance reasons. On top of that it will not work with k8s.
I agree after reading the full thread[0] that Bin share and reading docker documentation that yes swap is not working on k8, thanks for the clarification it helps me a lot. Now related to the performance. I do understand that if the scheduler sends a pod to a machine it should never use swap at all. but here my question ( as Linux performance eng ) could be. Why ? and what benchmarks prove that ?. As far as I know, please correct me if I am wrong, It’s normal and can be a good thing for Linux systems to use some swap, even if there is still available RAM. The Linux Kernel will move memory pages which are hardly ever used into swap space to ensure that even more cachable space is made available in-memory for more frequently used memory pages (a page is a piece of memory). Swap usage becomes a performance problem when the Kernel is pressured to continuously move memory pages in and out of memory and swap space. ( is this the corner case that we are preventing not to use the swap ) . Another advantage is that swap gives admins time to react to low memory issues. We will often notice the server acting slowly and upon login will notice heavy swapping. Without swap running out of memory can create much more sudden and severe chain reactions. Obed/Dean, do you know of any benchmark or use case that proves this ? just wondering [0] https://github.com/kubernetes/kubernetes/issues/53533
Regards
Victor R
On Mon, Apr 29, 2019 at 8:34 AM Miller, Frank <Frank.Miller@windriver.com> wrote:
We will be holding a meeting today. Planned agenda is below – please add to the etherpad agenda if you have any additional topics to discuss:
Etherpad: https://etherpad.openstack.org/p/stx-containerization
Agenda:
1. Sanity status/issue discussion:
- All in one systems, bare metal
- Multi node systems
- Virtual systems
2. Test team status & top issues
3. Feature Topics:
- Overall status
- Technical discussion on any outstanding features
4. Open topics
- no meeting Monday May 6th
- other items?
------
Frank
_______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss
_______________________________________________ Starlingx-discuss mailing list Starlingx-discuss@lists.starlingx.io http://lists.starlingx.io/cgi-bin/mailman/listinfo/starlingx-discuss