| opendevreview | Anishere Mariam proposed openstack/bifrost master: Add troubleshooting note for testenv after VM reboot https://review.opendev.org/c/openstack/bifrost/+/982361 | 06:37 |
|---|---|---|
| opendevreview | Abhishek Bongale proposed openstack/ironic-tempest-plugin master: Add conditional graphical console validation https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/982077 | 09:12 |
| opendevreview | Abhishek Bongale proposed openstack/ironic master: Limit interfaces to fake-graphical for testing https://review.opendev.org/c/openstack/ironic/+/982270 | 09:14 |
| opendevreview | Abhishek Bongale proposed openstack/ironic master: Limit interfaces to fake-graphical for testing https://review.opendev.org/c/openstack/ironic/+/982270 | 09:36 |
| cid | Friday Ironic, o/ | 10:24 |
| cid | cardoe, JayF, that makes total sense as natural v2 for runbooks | 10:25 |
| cid | May or may not even need a spec. I guess it depends on. | 10:27 |
| cid | if there’s an immediate edge case . | 10:28 |
| opendevreview | Harald Jensås proposed openstack/networking-generic-switch master: Fix Arista EOS vxlan vlan unplug syntax https://review.opendev.org/c/openstack/networking-generic-switch/+/982392 | 11:11 |
| opendevreview | Harald Jensås proposed openstack/networking-generic-switch master: Fix Arista EOS route-target auto syntax https://review.opendev.org/c/openstack/networking-generic-switch/+/982393 | 11:11 |
| opendevreview | Harald Jensås proposed openstack/networking-baremetal master: Remove redundant IDL table existence checks https://review.opendev.org/c/openstack/networking-baremetal/+/982342 | 11:34 |
| opendevreview | Mahnoor Asghar proposed openstack/ironic master: Soft power off before ejecting virtual media regardless of CONF.inspector.power_off https://review.opendev.org/c/openstack/ironic/+/981942 | 11:50 |
| opendevreview | Harald Jensås proposed openstack/networking-generic-switch master: Arista EOS: Create VLAN before VXLAN mapping https://review.opendev.org/c/openstack/networking-generic-switch/+/982400 | 12:15 |
| opendevreview | Mahnoor Asghar proposed openstack/ironic master: Soft power off before ejecting virtual media regardless of CONF.inspector.power_off https://review.opendev.org/c/openstack/ironic/+/981942 | 12:31 |
| *** alegacy__ is now known as alegacy | 12:43 | |
| JayF | cid: because it would be a change to the API, it would need a spec even if just a small one | 12:45 |
| cid | Oooh, okay. | 12:46 |
| TheJulia | good morning | 13:23 |
| opendevreview | Harald Jensås proposed openstack/networking-generic-switch master: Add redistribute learned to Arista EOS L2VNI conf https://review.opendev.org/c/openstack/networking-generic-switch/+/982424 | 13:56 |
| cardoe | TheJulia: once all of hjensas's changes land are we gonna make a follow on release of N-B and N-G-S? | 14:56 |
| TheJulia | As long as none of them are super featurey we can backport them | 14:59 |
| TheJulia | and then cut another release from tha tbranch | 14:59 |
| TheJulia | but we'll be held to *after* R-0 | 14:59 |
| TheJulia | (marketing (for some reason) doesn't want any package version numbers changing) | 15:00 |
| TheJulia | (At least, that is what I've been told in the past) | 15:00 |
| opendevreview | Doug Goldstein proposed openstack/ironic-specs master: Runbooks with Multiple Traits https://review.opendev.org/c/openstack/ironic-specs/+/982441 | 15:11 |
| cardoe | cid / JayF: ^ | 15:11 |
| hjensas | TheJulia: With arista, I now see arp requests from the ironic node, but afict it is dropped by ovs. | 15:42 |
| hjensas | 15:39:07.583045 fa:16:3e:06:93:f8 > fa:16:3e:c3:ba:c5, ethertype 802.1Q (0x8100), length 64: vlan 105, p 0, ethertype ARP (0x0806), Ethernet (len 6), IPv4 (len 4), Request who-has 10.0.5.1 tell 10.0.5.65, length 46 | 15:42 |
| hjensas | match=(inport == "neutron-d9ab994d-a00b-48d1-9dea-ec3b0244a8ac-localnet-public" | 15:43 |
| hjensas | && eth.src == fa:16:3e:06:93:f8 | 15:43 |
| hjensas | && eth.dst == fa:16:3e:c3:ba:c5 | 15:43 |
| hjensas | && !is_chassis_resident("a52fafc3-b09f-498e-a9dc-90a71156bfd8")), | 15:43 |
| hjensas | action=(drop;) | 15:43 |
| TheJulia | .... | 15:43 |
| TheJulia | so mentally unwinding it, we swap from one type to another type of switch, change nothing else ml2 wise, and we we're now dropping packets in OVS on the way into OVN? | 15:44 |
| TheJulia | on the networker node? | 15:44 |
| hjensas | I think I need to spin up the nxos lab in parallel and see if that has the same rule. | 15:44 |
| hjensas | yeah, I wonder if supress-arp which I added on NXOS is why it can work anyway. Will set up the nxos lab side by side to compare, and start crawling Arista docs on ARP suppress - there is no similar command afict. | 15:46 |
| TheJulia | I mean, the ml2 in ngs shouldn't be causing that at all, I'd suspect you'd get a similar rule. | 15:46 |
| TheJulia | err, joy | 15:47 |
| TheJulia | hmm | 15:47 |
| TheJulia | It *should* *really* be trying to use the local switch arp cache and ingress-replication to discover the addresses, I guess the issue with that rule is it doesn't allow arp to be resolvable, ever | 15:47 |
| TheJulia | which itself is super weird | 15:48 |
| TheJulia | in other words, its likely fine to send out the arp and for OVS to drop it as long as something is able to get and respond to it, but it sounds like thats fundimentally broken on the physical attachment point. Maybe we need to see if we can explicitly include arp.. somehow.?! | 15:48 |
| hjensas | I guess it is trying to protect against arp spoofing. | 15:48 |
| TheJulia | yeah | 15:49 |
| TheJulia | I still don't see how it would have worked with cisco and not for arista unless arista is just entirely disqualifying upon a lack of arp response | 15:55 |
| TheJulia | OH! look at the match and the rule. I wonder if somehow the chassis is mismatched? or somehow its getting considered on a different port? | 15:56 |
| opendevreview | Andrew Timmes proposed openstack/ironic-python-agent master: Add bootaa64.csv to BOOTLOADERS_EFI https://review.opendev.org/c/openstack/ironic-python-agent/+/982470 | 17:06 |
| opendevreview | Andrew Timmes proposed openstack/ironic-python-agent master: Add bootaa64.csv to BOOTLOADERS_EFI https://review.opendev.org/c/openstack/ironic-python-agent/+/982470 | 17:11 |
| opendevreview | Adam Harwell proposed openstack/sushy stable/2026.1: Prevent double-wrapping timeout tuple in Connector _op https://review.opendev.org/c/openstack/sushy/+/982471 | 17:18 |
| rm_work | ^^ i think that's an easy backport? | 17:22 |
| rm_work | I am trying to remember what the other thing I was trying to get backported was... but we've switched to running on 2026.1 at this point | 17:22 |
| opendevreview | Syed Haseeb Ahmed proposed openstack/ironic master: Refactor idrac-redfish SKU→serial_number override to run before inventory storage https://review.opendev.org/c/openstack/ironic/+/982478 | 17:38 |
| TheJulia | Yeah, likely easy, I'm sort of blinking though in the confused way of "how did you get there?" | 17:45 |
| rm_work | lol | 17:46 |
| rm_work | so, TBH you may have noticed, I do not always have THE BEST idea what is happening here... what I'm finding myself mostly doing is helping manage patches/build infra for people who are doing the actual low level debugging 😆 | 17:46 |
| rm_work | they tell me we need a patch, I make sure they have the patch and try to track it through to merging while managing it downstream temporarily and giving them builds 🙂 | 17:47 |
| rm_work | they do often explain the issue to me though so I try to have at least a cursory level of understanding about why we need certain patches, but in this case, I actually don't know what they ran into | 17:49 |
| opendevreview | Syed Haseeb Ahmed proposed openstack/ironic master: Fix idrac-redfish SKU→serial_number override to run before rule evaluation https://review.opendev.org/c/openstack/ironic/+/982478 | 18:02 |
| TheJulia | rm_work: that is totally fair! :) | 18:13 |
| hjensas | table=26(ls_in_external_port), priority=100 , match=(inport == "neutron-271c03fe-a2da-443a-9743-a0d6331bfee0-localnet-public" && eth.src == fa:16:3e:fb:e8:0b && eth.dst == fa:16:3e:41:b2:f2 && !is_chassis_resident("265d49b4-269e-45bd-afc3-f22599fc5db8")), action=(drop;) | 19:15 |
| hjensas | table=26(ls_in_external_port), priority=100 , match=(inport == "neutron-d9ab994d-a00b-48d1-9dea-ec3b0244a8ac-localnet-public" && eth.src == fa:16:3e:06:93:f8 && eth.dst == fa:16:3e:c3:ba:c5 && !is_chassis_resident("2ff6448d-4444-4f4d-b6ae-d885dc2d7ff4")), action=(drop;) | 19:15 |
| hjensas | TheJulia: yeah, identical rule on the nxos env. :-| | 19:16 |
| TheJulia | hjensas: is it just arp handling differences on the l2vni ? or is it all verifiable packet handling differences? I guess the huge unknown is why does one work and the other not work with the same rule | 19:53 |
| hjensas | I think it is the arp supression, the router sends garp so leaf01 can learn the mac+ip and when the client does arp it can learn it's mac+ip. And arp supression is then allowing the fabric to just reply. | 19:56 |
| hjensas | There is this in Arista docs: Type-2 Route: Host advertisement Route | 19:57 |
| hjensas | Used to advertise the reachability of a MAC address, or optionally a MAC and IP binding as learned by a specific EVI. With the advertisement of the optional IP address of the host, EVPN provides the ability for VTEPs to perform arp suppression and arp proxy to reduce flooding within the Layer 2 VPN. | 19:57 |
| hjensas | Trying to figure out how to turn it on ... | 19:57 |
| TheJulia | Hmm, multiswitch in that case, or is this even with single switch? | 20:40 |
| opendevreview | Julia Kreger proposed openstack/ironic-python-agent master: WIP: Enforce TLS 1.2 minimum across all connections https://review.opendev.org/c/openstack/ironic-python-agent/+/982174 | 20:45 |
| opendevreview | Julia Kreger proposed openstack/ironic-python-agent master: WIP: Upgrade TLS certificates to P-384 curve with configurable support https://review.opendev.org/c/openstack/ironic-python-agent/+/982175 | 20:45 |
Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!