November 2021 - openstack-discuss

[tacker][ptg] Yoga PTG summary
by yasufum 01 Nov '21

01 Nov '21

Hi everyone, Thank you for participating in the Yoga PTG sessions. We discussed 17 items totally, including 13 specs, and agreed all to implement the proposed features. We also decided to set the spec freeze on 31th Dec. The etherpad of our PTG is at https://etherpad.opendev.org/p/tacker-yoga-ptg Here is a summary of the PTG sessions. *Day 1 1. Introduce for Tacker installer * To reduce uinteresting steps of setting up tacker, we introduce a dedicated installer for developers. * For beginners, we should minimize difficulties of deploying and make them focusing on their usecases. 2. Prometheus monitoring and AutoHeal support for Kubernetes Cluster VNF via FM Interface * Add fault management interface support applied with ETSI SOL003 standards including polling mode and notify mode. * Polling mode responds to resources that is named "Alarms" and "Individual alarm". On the other hand, notify mode responds to resources that is named "Subscriptions" and "Notification endpoint" 3. Support VNF update operations reflecting changes to VNF instances using MgmtDriver * Add update APIs for container VNF instances by using ConfigMap and Secret of k8s. 4. Support CentOS stream * Revise our current incomplete CentOS support for the latest 'stream'. * We will provide some functional tests for the update, but non-voting for a while. 5. Add VNF Package sample for practical use cases * Provide more practical examples of usecase of tacker for users because current usecases in documentation are simple and not enough so much for actual cases. * There are three examples supposed in the proposal, "use multiple deployment flavours", "deploy VNF connected to external network" and "deploy VNF as HA cluster". *Day 2 6. Support handling large query results by ETSI NFV * Introduce an extended feature for reducing a large amount of result for some queries defined in ETSI standards. * The idea is simply dividing the results in several pieces with paging. 7. Support FT of Multi tenant * There are two problems in multi-tenancy case, (1) No restriction in assigning a tenant,(2) Notifying events on different tenants. * The policy of assigning a tenant should be clarified and also Notification process under multi-tenancy should be changed. * We need to clear the policy of RBAC (Is it allowed to make admin user can access all resources?) 8. Sample Ansible Driver * There are no management drivers enable VNF Configuration using Ansible. So, add the scripts as samples and docs. * Revise the name of directories for considering conventioins in takcer repo is necessary. * This can be used for 'Add VNF Package sample for practical use cases' proposal. We wii support the sample development. 9. Support Robot Framework API * Currently Tacker functional tests mainly focus on checking various VNF patterns such as a simple VNF, multi VDU, volume attach and affinity set. Tacker community is advancing ETSI NFV standard compliance, and coverage of compliant API testing becomes important. * The proposal is to use Robot Framework to achieve automated API testing and in the first step, we adopt API test code released by ETSI NFV-TST010. 10. Add Tacker-horizon Unit Test Cases * `tacker-horizon` which provides very limited features, such as showing a list of registered VIMs, VNFDs or so. We don't use this feature so much without checking instances of Tacker quickly via intuitive Web GUI way. So, we have not maintained tacker-horizon so actively * One of the reasons why bugs in tacker-horizon have not fixed is that we have no unittests in. Although we can find a bug coincidentally while using the features, but should implement unittests because bugs in horizon are some contextual and not so easy to find by hand. *Day 3 11. Report for ETSI NFV API Conformance Test Results * Share the result of Remote NFV&MEC API Plugtest 2021, Totally 136 API Conformance test sessions were executed for [NFV-SOL003] were executed as well as for NFV-SOL005. * https://hackmd.io/q4DzQ6_2Q0e-TdmtBikVlQ?view 12. Reduce code clone from sol-kubernetes job of FT * There are much similar files can be reduced for functional tests. It makes maintenance more complicated and difficult. The goal is to reduce it to less than 20%. * In particular, since sol_kubernetes's code clone rate is up to 40% , we think it will be better to refactor to make future maintenance easier. * other details: https://hackmd.io/Wo8cBIH_RPe6ll1hNwmx1w?view 13. Reduce unnecessary test codes * It's similar as above item. We have very similar YAML files for definitions and useless template files for tests can be reduced. 14. Enhance NFV SOL_v3 LCM operation * Introduce the latest V2 APIs for LCM operation as below. * Scale VNF (POST /vnf_instances/{vnfInstanceId}/scale) * Heal VNF (POST /vnf_instances/{vnfInstanceId}/heal) * Modify VNF Information (PATCH /vnf_instances/{vnfInstanceId}) * Change External VNF Connectivity (POST /vnf_instances/{vnfInstanceId}/change_ext_conn) 15. Support ETSI NFV-SOL_v3 based error-handling operation * Introduce the latest V2 APIs for error handling operation as below. * Retry operation (POST /vnf_lcm_op_occs/{vnfLcmOpOccId}/retry) * Fail operation (POST /vnf_lcm_op_occs/{vnfLcmOpOccId}/fail) * Rollback operation (POST /vnf_lcm_op_occs/{vnfLcmOpOccId}/rollback) * Test scenario needs to include raising an error, becoming FAILD_TEMP, and executing the ErrorHandling API. * Timer adjusting and using inapporpriate vnf package can cause error. 16. Support ChangeCurrentVNFPackage * Add API for ChangeCurrentVNFPackage by which blue-green deployment and rolling update are supported. * Both of VIMs, OpenStack and Kubernetes, are covered. 17. Support heal and scale method in lcm_user_data * Enable to customize stack parameters for heal and scale operations in user script, `user_lcm_data.py` more specifically. * Call the proposed methods if it's existing in the script for the operations. Thanks, Yasufumi

1 0

[horizon][dev]Handle multiple login sessions from same user in Horizon
by Luz de Avila, Arthur 01 Nov '21

01 Nov '21

Hi everyone, In order to improve the system my colleagues and I would like to bring up a new feature to Horizon. We found out that an user is able to login in Horizon with the same credentials in multiple devices or/and browsers. This may be not very secure as the user can login in many different devices or/and browsers with the same credential. Thinking on that, we would like bring up more control to the admin of the system in a way that the admin can enable or disable the multiple login sessions according to the needs of the system. For a better follow up of this propose, a blueprint has been opened with more details about the idea and concepts of this and we would like the onion of the community whether this feature make sense to implement or not. The blueprint opened on launchpad: https://blueprints.launchpad.net/horizon/+spec/handle-multiple-login-sessio… Kind regards, Arthur Avila

2 1

[charms] Yoga PTG Summary
by Alex Kavanagh 01 Nov '21

01 Nov '21

Hi All Thanks to all who participated in the PTG discussions. We had some good sessions where we tackled thorny issues in the charms deployments of OpenStack. A brief set of highlights is: - The store holding all the charms (https://jaas.ai/) just happens to be migrating to a new home (https://charmhub.io/) within the timescales of the Yoga cycle and some of the changes in behaviour provide some challenges and also opportunities to both simplify the charms and provide a better operator experience. The discussions highlighted the pain points and the general direction that the migration could take around: - Charm upgrades - OpenStack upgrades - Series upgrades of the underlying operating system. - The charms deployment of OpenStack supports versions back to mitaka (pre yoga) and queens (yoga cycle onwards). This has recently presented challenges around Python 3 support, particularly with py35 being EOL September 2020, and with py36 (bionic LTS) EOL 23rd December 2021. We discussed various strategies and approaches around maintenance, security and support of queens onwards in terms of support and upgrades. - The charms team also maintains a CI infrastructure based on Jenkins and (more recently) self-hosted Zuul to test installation and upgrade of OpenStack solutions. We discussed the further migrations of services from Jenkins to Zuul and issues associated with it. - Discussions around building new charms and refreshing old charms for the new components and features in the previous xena cycle and oncoming yoga cycle. Our raw notes are here: https://etherpad.opendev.org/p/charms-yoga-ptg Thank you again for all the participation in setting the direction and priorities for the yoga cycle. -- Alex Kavanagh - PTL Yoga

1 0

Openstack ansible VS kolla ansible
by A Monster 01 Nov '21

01 Nov '21

Openstack-ansible uses LXC containers to deploy openstack services , while Kolla uses docker containers instead, which of these two deployment tools should I use for an Openstack deployment, and what are the differences between them.

3 2

is OVS+DPDK useful for general purpose workload
by Satish Patel 01 Nov '21

01 Nov '21

Folks, I have deployed openstack and configured OVS-DPDK on compute nodes for high performance networking. My workload is general purpose workload like running haproxy, mysql, apache and XMPP etc. When I did load testing I found performance was average and after 200kpps packet rate I noticed packet drops. I heard and read that DPDK can handle millions of packets but in my case its not true. I am using virtio-net in guest vm which processes packets in the kernel so I believe my bottleneck is my guest VM. I don't have any guest based DPDK applications like testpmd etc. does that mean OVS+DPDK isn't useful for my cloud? How do I take advantage of OVS+DPDK with general purpose workload? Maybe I have the wrong understanding about DPDK so please help me :) Thanks ~S

2 2

Re: Using ceph for openstack storage
by A Monster 01 Nov '21

01 Nov '21

Thank you for your response. Sadly, I'm talking about actual production, but I'm very limited in terms of hardware. I was thinking about using RAID for controller node as data redundancy, because I had the idea of maximizing the number of nova compute nodes, So basically i thought off using a controller with the following services ( Nova, Neutron, Keystone, Horizon, Glance, Cinder and Swift). Following the configuration you suggested, I would have : - 3 Controllers that are also Ceph Monitors - 9 Nova compute nodes and Ceph OSDs My questions are : - having multiple Ceph monitors is for the sake of redundancy or does it have a performance goal ? - Combining Ceph OSD and Nova compute wouldn't have performance drawbacks or damage the integrity of data stored in each node. wouldn't it be better is this case to use two separate servers for swift and glance and use RAID for data redundancy instead of using Ceph SDS. Thank you very much for you time.

3 2

Re: Router interfaces are down
by Slawek Kaplonski 01 Nov '21

01 Nov '21

Hi, Please don't drop ML from the thread. You have to go to the node where Your router is hosted and investigate there in the agent's logs. If ports are DOWN, I would start with checking in the L2 agent logs (neutron-ovs-agent or linuxbridge-agent, idk what You are using exactly). If there is no any errors there, You can also check neutron-server logs, why ports aren't set to be UP as well as checking neutron-l3-agent logs on the node where router is hosted. On czwartek, 28 października 2021 18:41:29 CET Jibsan Joel Rosa Toirac wrote: > Here I let you some screenshots of my Network Topology: > > > > El jue, 28 oct 2021 a las 7:43, Jibsan Joel Rosa Toirac (<jibsan94(a)gmail.com>) > escribió: > > Yes it’s neutron router. Well the router is centralized. It is located > > between the subnets and the node, all the subnets will pass through the > > router to internet, but I don’t know what else check to set it up for. I’m > > using a friend’s Openstack instance on a server to check out if I’m missing > > something and both nodes are the same. I will send you a screenshot of my > > network Topology. > > > > Greetings > > > > On Thu, Oct 28, 2021 at 2:33 AM Slawek Kaplonski <skaplons(a)redhat.com> > > > > wrote: > >> Hi, > >> > >> On środa, 27 października 2021 21:39:55 CEST Jibsan Joel Rosa Toirac > >> > >> wrote: > >> > Hello, I'm trying to route all the requests from a private vlan to > >> > internet. I have a private network and all the Virtual Machines inside > >> > >> the > >> > >> > subnet config can do everything between they, but if I ping to > >> > >> Internet, it > >> > >> > doesn't work. > >> > > >> > When I see the router_external it says all the interfaces are DOWN. > >> > >> By router_external, You mean neutron router, right? > >> If so, what kind of router it is, centralized HA or non-HA, or maybe DVR? > >> Is router scheduled properly to some node? You can check that with > >> command > >> like "neutron l3-agent-list-hosting-router <router_id>". > >> > >> > I have search in everywhere but I can't find a solution for this. > >> > > >> > Thank you for your time > >> > >> -- > >> Slawek Kaplonski > >> Principal Software Engineer > >> Red Hat -- Slawek Kaplonski Principal Software Engineer Red Hat

1 0

[Diversity] Diversity and Inclusion Meeting Reminder
by Amy Marrich 31 Oct '21

31 Oct '21

The Diversity & Inclusion WG invites members of all OIF projects to attend our next meeting Monday November 1st, at 17:00 UTC in the #openinfra- diversity channel on OFTC. The agenda can be found at https://etherpad.openstack.org/p/diversity-wg-agenda. Please feel free to add any topics you wish to discuss at the meeting. Thanks, Amy (apotz)

1 0