| *** amoralej|off is now known as amoralej | 06:53 | |
| zigo | frickler: https://review.opendev.org/c/openstack/neutron-dynamic-routing/+/849510 didn't help and I still get my retry thingy to do its work ... :/ | 07:59 |
|---|---|---|
| frickler | zigo: did you test on the agent or on the controller? from what you said yesterday, I'm assuming agent, but now it occurs to me that the changed code is actually running on the controller. btw. do you have dedicated neutron-rpc-server processes deployed? | 08:32 |
| opendevreview | Rodolfo Alonso proposed openstack/neutron stable/xena: Implement specific tracked resource count method per quota driver https://review.opendev.org/c/openstack/neutron/+/851918 | 08:46 |
| opendevreview | wangjiaqi proposed openstack/neutron-vpnaas-dashboard master: remove unicode from code https://review.opendev.org/c/openstack/neutron-vpnaas-dashboard/+/851994 | 11:17 |
| opendevreview | Vasyl Saienko proposed openstack/neutron-lib master: Update meaning for router:external https://review.opendev.org/c/openstack/neutron-lib/+/851981 | 11:25 |
| bcafarel | haleyb: yeah I use https://gist.github.com/yoctozepto/7ea1271c299d143388b7c1b1802ee75e/raw/7ad5a5639ee58c7d2ec3bdfd46e017d19bee5b92/gerrit_3_zuul_status.user.js but it does need update for recent zuul | 12:13 |
| bcafarel | zuul updates are not fully pasted as comments anymore, you need to fetch from zuul.openstack.org - probably not too complicated but you need some JS expertise :) | 12:13 |
| bcafarel | haleyb: keep me posted if you find a working update version :) | 12:17 |
| *** amoralej is now known as amoralej|lunch | 12:28 | |
| *** dasm|off is now known as dasm | 13:18 | |
| *** amoralej|lunch is now known as amoralej | 13:29 | |
| *** lifeless_ is now known as lifeless | 13:51 | |
| haleyb | bcafarel: yes, that's the script i use too, and i have no JS experience so even debugging it isn't going to happen :( | 13:56 |
| ralonsoh | haleyb, hey! do you know why if we stop the radvd in the controller that has the active router namespace, a VM IPv6 traffic (ping traffic) is stopped? | 14:07 |
| opendevreview | Merged openstack/neutron stable/train: Port update will trigger less notifications to the DHCP agents https://review.opendev.org/c/openstack/neutron/+/851530 | 14:09 |
| haleyb | ralonsoh: is it losing a route? i wonder if it sends an RA when it's stopping to remove itself as the default router? | 14:09 |
| ralonsoh | haleyb, uhhh I need to check that | 14:10 |
| ralonsoh | good advise | 14:10 |
| haleyb | ralonsoh: check link-local first, that should always work, anything global will require a default route | 14:11 |
| ralonsoh | haleyb, right, I would also check the routing table inside the VM | 14:11 |
| haleyb | ralonsoh: yup, my guess is either a route or address being removed is causing it, been a while since i looked at radvd startup/shutdown | 14:13 |
| ralonsoh | yeah, because the service itself is not routing the traffic | 14:14 |
| mtomaska | haleyb: ralonsoh: I am not sure, I will look at tcpdump inside router namespace | 14:17 |
| haleyb | running tcpdump/wireshark might help too if radvd is sending something at shutdown | 14:17 |
| mtomaska | :) | 14:17 |
| haleyb | yes, message collision :) | 14:17 |
| mtomaska | router default is a flag inside RA message right? I will have to look up packet details | 14:18 |
| haleyb | yes, i believe it's a single flag, source address of RA will be the default router if set | 14:19 |
| haleyb | been a while since i've looked at the RFCs | 14:19 |
| haleyb | So the RA has a "Router Lifetime" field - if 0 indicates should not be a default router, else it's the number of seconds the sender should be the default router | 14:24 |
| haleyb | https://datatracker.ietf.org/doc/html/rfc4861 | 14:24 |
| *** amoralej is now known as amoralej|off | 15:21 | |
| opendevreview | Vasyl Saienko proposed openstack/neutron-lib master: Update meaning for router:external https://review.opendev.org/c/openstack/neutron-lib/+/851981 | 16:21 |
| mtomaska | haleyb: During normal operation a RA packet coming from qr route interface contains Router Lifetime(s) 300. When radvd daemon is shutdown there is one more RA packet containing Router Lifetime(s) : 0. Then naturally, VM ip -6 neigh for that route goes STALE | 16:38 |
| mtomaska | no wait | 16:39 |
| haleyb | mtomaska: thanks for the info, and i guess expected - any VM receiving that should remove their default route via that address | 16:39 |
| frickler | under what condition does the radvd get stopped? | 16:39 |
| mtomaska | I see VM ip -6 route going back to REACHABLE | 16:40 |
| mtomaska | weird... let me run tcpdump on VM to see if there are some packets coming in | 16:40 |
| mtomaska | fricker: currently when l3-agent is stopped it stops radvd. I will have to look into l3-agent code to see when that get fired off | 16:41 |
| frickler | maybe one should change the whole active/passive concept for IPv6 HA, instead have multiple active (can even be >2) instances. (just thinking out loud before EODing) | 17:25 |
| *** EugenMayer4 is now known as EugenMayer | 21:39 | |
| opendevreview | Miguel Lavalle proposed openstack/neutron master: [DNM][POC]Bump revision version when updating description https://review.opendev.org/c/openstack/neutron/+/852053 | 22:58 |
| *** dasm is now known as dasm|off | 23:26 | |
| opendevreview | ayenachew molla proposed openstack/neutron-tempest-plugin master: Delete tcp rule from a security group https://review.opendev.org/c/openstack/neutron-tempest-plugin/+/848268 | 23:30 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!