openstack-discuss

openstack-discuss@lists.openstack.org

December 2020

  • 145 participants
  • 188 discussions
Fwd: DPDK+OVS with OpenStack
by Mark Wittling 23 Apr '25

23 Apr '25
Looking for someone who knows OpenStack with OpenVSwitch, and in additionto that, DPDK with OpenStack and OVS.I am using OpenStack Queens, withOpenVSwitch. The architecture I am using is documented here:https://docs.openstack.org/neutron/queens/admin/deploy-ovs-provider.htmlTheOVS I am using on the Compute Node, is compiled with DPDK, and I haveenabled the datapath to netdev (DPDK) on br-prv (provider network bridge),and br-tun (tunneling bridge). But these two bridges, br-tun and br-prv,are patched into another OpenStack bridge, called br-int. I wasn’t actuallysure about whether to tinker with this bridge, and wondered what datapathit was using.Then, I realized there is a parameter in theopenvswitch_agent.ini file, which I will list here:# OVS datapath to use. 'system' is the default value and corresponds to the# kernel datapath. To enable the userspace datapath set this value to 'netdev'.# (string value)# Possible values:# system - <No description provided># netdev - <No description provided>#datapath_type = systemdatapath_type = netdevSo in tinkering with this, what I realized, is that when you set thisdatapath_type to system or netdev, it will adjust the br-int bridge to thatdatapath type.So here is my question. How can I launch a non-DPDK VM, ifall of the bridges are using the netdev datapath type?Here is anotherquestion. What if one of the flavors don’t have the largepages property seton them? I assumed OpenStack would revert to a system datapath and not useDPDK for those VM interfaces. Well, I found out in testing, that is not thecase. If you set all your bridges up for netdev, and you don’t set theproperty on the Flavor of the VM (largepages), the VM will launch, but itsimply won’t work.Is there no way, on a specific Compute Host, to supportboth DPDK (netdev datapaths) and non-DPDK (system datapaths)?Either on a VMinterface level (VM has one interface that is netdev DPDK and another thatis system datapath non-DPDK)?Or on a VM by VM basis (VM 1 has 1 or morenetdev datapath interfaces and VM 2 has 1 or more system datapathinterfaces)?Am I right here? Once you set up a Compute Host for DPDK, it’sDPDK or nothing on that Compute Host? (edited)
3 2
0 0
How to config DPDK compute in OVN
by dangerzone ar 22 Apr '25

22 Apr '25
Hi....as far I understand ovs does support dpdk compute to be deployed. I'mreferring to the guideline '*https://docs.openstack.org/networking-ovn/queens/admin/dpdk.html<https://docs.openstack.org/networking-ovn/queens/admin/dpdk.html>* ' ...It's too brief.. May I know how to config dpdk compute in OVN networkingopenstack? Hope someone could help what is the config to be done....Currently compute is run without dpdk over openstack train ovn based.Please need some info on how I can proceed. Thank you
2 1
0 0
[all] Dynamic Zuul results table in Gerrit 3
by Radosław Piliszek 04 Aug '22

04 Aug '22
Hello Fellow OpenStack and OpenDev Folks!TL;DR click on [3] and enjoy.I am starting this thread to not hijack the discussion happening on [1].First of all, I would like to thank gibi (Balazs Gibizer) for hackinga way to get the place to render the table in the first place (punintended).I have been a long-time-now user of [2].I have improved and customised it for myself but never really got toshare back the changes I made.The new Gerrit obviously broke the whole script so it was of no use toshare at that particular state.However, inspired by gibi's work, I decided to finally sit down andfix it to work with Gerrit 3 and here it comes: [3].Works well on Chrome with Tampermonkey. Not tested others.I hope you will enjoy this little helper (I do).I know the script looks super fugly but it generally boils down to amix of styles of 3 people and Gerrit having funky UI rendering.Finally, I'd also like to thank hrw (Marcin Juszkiewicz) for linkingme to the original Michel's script in 2019.[1]http://lists.openstack.org/pipermail/openstack-discuss/2020-November/019051…[2]https://opendev.org/x/coats/src/commit/444c95738677593dcfed0cfd9667d4c4f0d5…[3]https://gist.github.com/yoctozepto/7ea1271c299d143388b7c1b1802ee75eKind regards,-yoctozepto
5 7
0 0
All,A few years ago I asked a question[1] about why nova, when given a hw_architecture property from glance for an image, would not end up using the correct qemu-system-xx binary when starting the guest process on a compute node if that compute nodes architecture did not match the proposed guest architecture. As an example, if we had all x86 hosts, but wanted to run an emulated ppc guest, we should be able to do that given that at least one compute node had qemu-system-ppc already installed and libvirt was successfully reporting that as a supported architecture to nova. It seemed like a heavy lift at the time, so it was put on the back burner.I am now in a position to fund a contract developer to make this happen, so the question is: would this be a useful blueprint that would potentially be accepted? Most of the time when people want to run an emulated guest they would just nest it inside of an already running guest of the native architecture, but that severely limits observability and the task of managing any more than a handful of instances in this manner quickly becomes a tangled nightmare of networking, etc. I see real benefit in allowing this scenario to run natively so all of the tooling that exists for fleet management 'just works'. This would also be a significant differentiator for OpenStack as a whole.Thoughts?[1]http://lists.openstack.org/pipermail/openstack-operators/2018-August/015653…Chris ApseyDirector | Georgia Cyber RangeGEORGIA CYBER CENTER100 Grace Hopper Lane | Augusta, Georgia | 30901https://www.gacybercenter.org
2 2
0 0

04 Oct '21
Hi,As you may know, eventlet is incompatible with dnspython >= 2.0.0.0rc1.See [1] for the details. However, Debian unstable has 2.0.0.Would there be some good soul willing to help me fix this situation? Iwould need a patch to fix this, but I'm really not sure how to start.Cheers,Thomas Goirand (zigo)
8 15
0 0

27 May '21
Hi everyone,This has been discussed several times on this mailing list in the past, but we never got to actually pull the plug.Ask.openstack.org was launched in 2013. The reason for hosting our own setup was to be able to support multiple languages, while StackOverflow rejected our proposal to have our own openstack-branded StackExchange site. The Chinese ask.o.o side never really took off. The English side also never really worked perfectly (like email alerts are hopelessly broken), but we figured it would get better with time if a big community formed around it.Fast-forward to 2020 and the instance is lacking volunteers to help run it, while the code (and our customization of it) has become more complicated to maintain. It regularly fails one way or another, and questions there often go unanswered, making us look bad. Of the top 30 users, most have abandoned the platform since 2017, leaving only Bernd Bausch actively engaging and helping moderate questions lately. We have called for volunteers several times, but the offers for help never really materialized.At the same time, people are asking OpenStack questions on StackOverflow, and sometimes getting answers there[1]. The fragmentation of the "questions" space is not helping users getting good answers.I think it's time to pull the plug, makeask.openstack.org read-only (so that links to old answers are not lost) and redirect users to the mailing-list and the "OpenStack" tag on StackOverflow. I picked StackOverflow since it seems to have the most openstack questions (2,574 on SO, 76 on SuperUser and 430 on ServerFault).We discussed that option several times, but I now proposed a change to actually make it happen:https://review.opendev.org/#/c/746497/It's always a difficult decision to make to kill a resource, but I feel like in this case, consolidation and simplification would help.Thoughts, comments?[1]https://stackoverflow.com/questions/tagged/openstack-- Thierry
10 18
0 0
[stein][neutron] gratuitous arp
by Ignazio Cassano 13 May '21

13 May '21
Hello, I have a problem on stein neutron. When a vm migrate from one nodeto another I cannot ping it for several minutes. If in the vm I put ascript that ping the gateway continously, the live migration works fine andI can ping it. Why this happens ? I read something about gratuitous arp.Please, help me ?Any workaround , please ?Best RegardsIgnazio
3 33
0 0
Hi all,Following up to the previous mailing list [1] and virtual meetup [2]discussions, I would like to announce the plans to deprecate the 'iscsi'deploy interface.This is the updated plan discussed on the virtual meetup:1) In the Victoria cycle (i.e. right now): - Fill in the detected feature gaps [3]. - Switch off the iscsi deploy interface by default. - Change [agent]image_dowload_source to HTTP by default. - Give the direct deploy a higher priority, so that it's used by defaultunless disabled. - Mark it as deprecated in the code (causing warnings when enabled). - Release a major version of ironic to highlight the defaults changes.2) In the W cycle: - Keep the iscsi deploy deprecated. - Listen to operators' feedback.3) In the X cycle - Remove the iscsi deploy completely from ironic and IPA. - Remove support code from ironic-lib with a major version bump.Please let us know if you have any questions or concerns.Dmitry[1]http://lists.openstack.org/pipermail/openstack-discuss/2020-August/016681.h…[2]https://etherpad.opendev.org/p/Ironic-Victoria-midcycle[3]https://storyboard.openstack.org/#!/story/2008075-- Red Hat GmbH,https://de.redhat.com/ , Registered seat: Grasbrunn,Commercial register: Amtsgericht Muenchen, HRB 153243,Managing Directors: Charles Cachera, Brian Klemm, Laurie Krebs, MichaelO'Neill
1 1
0 0
Hello Barbican team,Apologies for not including barbican in the previous thread on this topic:http://lists.openstack.org/pipermail/openstack-discuss/2020-December/019430…The situation is that cursive is used by Nova, Glance, and Cinder and we'd like to move it out of the 'x' namespace into openstack governance. The question is then what team would oversee it. It seems like a good fit for Oslo, and the Oslo team seems OK with that, but since barbican-core is currently included in cursive-core, it make sense to give the Barbican team first dibs. From the consuming teams' side, I don't think we have a preference as long as it's clear who we need to bother about approvals if a bugfix is posted for review.Thus my ask is that the Barbican team indicate whether they'd like to move cursive to the 'openstack' namespace under their governance, or whether they'd prefer Oslo to oversee the library.Thank you!brian
5 5
0 0
Hi Ian,Given our timezone difference, I've decided to send you an email, addingopenstack-discuss in CC for greater exposure.We're trying to make IPA (our agent ramdisk) to work without relying onDHCP for cases like Edge deployments. We've settled on thenetwork_data.json format for the API side and wanted to use Glean on theramdisk to apply it. You can read more details in [1].The problem is, I cannot make Glean work with any ramdisk I build. The cruxof the problem seems to be that NetworkManager (used by default in RHEL,CentOS, Fedora and Debian at least) starts very early, creates the defaultconnection and ignores whatever files Glean happens to write afterwards. OnDebian running `systemctl restart networking` actually helped to pick thenew configuration, but I'm not sure we want to do that in Glean. I haven'tbeen able to make NetworkManager pick up the changes on RH systems so far.I build ramdisks using IPA-builder [2] by adding the simple-init element.I've tried removing dhcp-all-interfaces (which we depend on by default) tono effect. I've tried disabling the DHCP server, ended up with no IPconnectivity at all. I haven't tried to shutdown and restart a connectionas recommended in [3] since it's not trivial to do via SSH.Do you maybe have any hints how to proceed? I'd be curious to know howstatic IP assignment works in the infra setup. Do you have images withNetworkManager there? Do you use the simple-init element?Any help is very appreciated.Dmitry[1]https://specs.openstack.org/openstack/ironic-specs/specs/not-implemented/L3…[2]https://opendev.org/openstack/ironic-python-agent-builder[3]https://mail.gnome.org/archives/networkmanager-list/2014-January/msg00032.h…-- Red Hat GmbH,https://de.redhat.com/ , Registered seat: Grasbrunn,Commercial register: Amtsgericht Muenchen, HRB 153243,Managing Directors: Charles Cachera, Brian Klemm, Laurie Krebs, MichaelO'Neill
4 9
0 0