From containers to extreme testing, Melvin Hillsman reports on hot topics from the recent event in Mexico City.

image

MEXICO CITY — It was great spending time with the OpenStack community in Mexico City. Our host venue — Centro de Cultura Digital — did not disappoint with an interactive light wall, multi-purpose spaces and a theater seating up to 120 where we held our sessions.

As is normally the case with the Ops Meetup, we had a good mix of users and developers which made for very interesting and interactive discussions. A full list of the Etherpads used for the Meetup can be found here: https://etherpad.openstack.org/p/MEX-ops-meetup.

We asked some folks who moderated sessions to comment on what stood out during the discussions:

Chris Morgan of Bloomberg

Documentation of older versions of OpenStack
I run an older version of OpenStack (nine production-quality Liberty clusters) and I still need the documents for it, I wanted to share our problems and see who else is in the same boat. We have a full-time contractor working for Bloomberg from Canonical who investigated this and it turns out that the commonly given answer for old documents (“just build from tag”) doesn’t work well at all. I explained that we’re pushing for a better solution for the entire community and that work is still in flight. For example, there’s some acceptance of keeping documentation longer in future, but distinguishing it better from the latest via use of watermarks, for example. We may be able to keep the Mitaka onwards documents longer  — they haven’t been taken down yet.

Sampath Priyankara of NTT Communications

Extreme testing

Since this is a quite new project, it took me most of the session to explain why and what we are doing in extreme testing. We had a good discussion on what we should include in the testing and what other aspects to consider. I started a follow-up ML thread — http://lists.openstack.org/pipermail/openstack-dev/2017-August/121104.html — to continue this discussion and getting good response.

Robert Starmer of Kumulus Technologies

OpenStack on containers:
Folks are interested in seeing a Kubernetes solution, but in the group, most had only dabbled with some container ops, and saw the value in at least containerization. Some in the group were still happy with, and continuing with the LXC based approach (OpenStack Ansible, or similar), and others are moving on with Kolla (a subset have tried it, a smaller subset have it in production). All saw the value of this mechanism, and those that aren’t on containers would like better guidance on how to migrate from bare metal deploy to container deploy (which was another session)

Containers on OpenStack:
Still massive confusion on what networking to use, what works, how the pieces are all to come together. Most are not aware if their customers are deploying containers _on_ their OpenStack systems (i.e., there’s no integration between the two domains). Some are investigating how one might manage container deployments (specifically Kubernetes) on OpenStack, and what integration makes sense (without RBAC in K8S that maps to Keystone, that doesn’t integrate well), Cinder -> K8S PVs works, LBaaS->k8s loadbalancer works, but again, what’s the integration model (is there even LBaaS available, etc.) Magnum is a no-op for the group, Zun is outside of most of the groups scope/interest. Almost no one had even heard of Kuryr (and the “right” network model didn’t seem to include Neutron anyway, or so I understood from a couple folks in the group).

Shintaro Mizuno of NTT Communications

Telco/NFV

NTT and AT&T shared their experiences on operating VNF in production. Most of the challenges come from the fact that the VNF(i.e. vEPCs) are monster applications, which were brought straight from legacy hardware, and requires full hardware performance. Technical challenges are around SR-IOV, OVS-DPDK, CPU pinning and NUMA placement which is to get performance VNF requires. If we have enough NFV operators in the future Meetups, we can discuss these issues in more detail. NTT is working on upstreaming NFV related features into Blazar (non-official project). Ops Telco/NFV WG is discussing to merge with LCOO WG.

Sean McGinnis of Huawei

TC feedback

I moderated to talk about the role of the Technical Committee and stimulate discussion around needs and issues. We answered some questions about what the TC does and does not do. The session ended up being mostly a discussion about having LTS releases. I really pushed that this is something that Red Hat, Canonical, Suse and Huawei need to collaborate to support. The biggest asks were to help skip level upgrades and longer support for releases.


These are just a few of the sessions we had during the Meetup. Overall, we had a great time and look forward to bringing these discussions to the community and furthering the work of OpenStack.

We greatly appreciate our sponsors Bloomberg, Centro de Cultura Digital, Nearsoft, Nearshore, Suse, Meltsan and Sentinel.

We look forward to seeing as many of you as we can during the next Meetup — March 7–8 2018 in Tokyo, Japan. Please remember to visit #openstack-operators IRC channel on Tuesdays @ 1400UTC and/or join the OpenStack Operators mailing list — http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators — to stay updated.

About the author

Melvin Hillsman is a self-taught tech enthusiast currently living in Houston, Texas. He has spent a number of years working as a Linux system administrator and now enjoys learning more about and working with open source communities. As a member of the OpenStack User Committee, he spends most of his days looking for opportunities to grow the user base of the OpenStack community. His love for technology, culture, and people has afforded him the opportunity to travel to many countries speaking about OpenStack and striving to understand and break down barriers to collaboration between those who develop open source software and those who use it.

 

Superuser wants to hear from operators for our ongoing interview series, get in touch at editorATopenstack.org.