2018.01.18 Specialty ERIS

Owned by Jamey McCabe
Last updated: Jan 18, 2018
3 min read

 at UTC 0300: | To see in your timezone and add to calendar: http://bit.ly/2mwebPA

Web meeting: https://zoom.us/j/579402487

 (add yourself to be notified of any changes)

Goals

  • finalize rollout/encourage more collaboration through SIG
  • plan any PTG attendance
  • coordinate on next steps

Discussion items

TimeItemWhoNotes
1Initial project status: We had determined in December meeting that we needed to get the initial project in place before inviting more collaborators.  Let's check on progress and enumerate remaining tasks.

Gautam started by sharing Lessons Learned

  1. Ansible playbook can burn 1 to 40 CPUs, all at once or in a chosen growth pattern. 
  2. Python mathematical function doesn't' use much CPU.  If you create a process that eats up memory you will get swapped out. 
  3. With Root you can create a routine to keep it from being swapped.In AT&T Systems we don't allow anyone or anything to have root.
  4. You may want stressNG installed but then need to install it on all machines and probably don't want this on production hardware.
  5. Have also created some Rally routines to create load
  6. They have 3 types of Runners,
    1. Serial - doesn't really generate load for us
    2. Constant
    3. Request per sec - preferred but need it for time not for script
  7. With Ansible it seems we do need a better orchestration tool to do things for certain events or tasks.  e.g. after CPU utilization reaches x, then kick off y.  May need an ERIS manager and an eventing service.  Trying to keep from bringing in software heavy solutions.

QA Spec - will be ready for Dublin.  Daisuke Nakazawa (Unlicensed) doing the updating and will join next meeting to status.

Sampath has installed ERIS in conjunction with the NTT Masakari CI environment.  This script integration can be shared.

2PTG Approach (Feb26-March 2, Dublin IR) - A working proposal was that from a SIGs and greater community perspective we join the [Openstack-sigs] [self-healing]  instead of having a separate [Openstack-sigs] [eris] SIG.  Some of that is dependent on the communities uptake not just our own decision.  Regardless it seems that the PTG conversations being setup by Adam Spiers may be a good entry for us if ready.  See links to the right.

https://www.openstack.org/ptg/

http://lists.openstack.org/pipermail/openstack-sigs/2018-January/000230.html

Seems likely we can get some hours during the first 2 days during QA time if we ask.  Sampath Priyankara agreed to follow up with QA team to ask.

Main action item is for us to check with our companies to see if they will send us.

2Overlap to Self Healing SIG.  Proposed that we just start participating there.
A good description of this related SIG, showing where the need for Destructive Testing is complimentary but would share and enhance efforts to bring out use cases. Basically each operator would bring their use case and be the testers of it.
The first meeting happened at the Denver PTG, and was minuted in this etherpad https://etherpad.openstack.org/p/self-healing-queens-ptg
The Sydney Forum pad contains an architecture diagram and list of user stories https://etherpad.openstack.org/p/self-healing-rocky-forum
4Action Items from previous meeting