| *** jph0 is now known as jph | 08:25 | |
| tkajinam | wondering if anyone has idea about this ... based on https://review.opendev.org/c/openstack/neutron/+/823637 now I'm setting rpc_worker=0 in ml2 ovn tests but I noticed there are still two RpcWorkers in ps | 08:46 |
|---|---|---|
| tkajinam | https://zuul.opendev.org/t/openstack/build/719a7ec5ca4b40b4a91bb88501233cc9 | 08:46 |
| tkajinam | neutron.conf is found in logs/etc/neutron and this contains rpc_workers=0, but ps result in logs/ps.txt shows two RpcWorker | 08:47 |
| tkajinam | neutron 70440 70395 70440 0.2 1.8 140796 150192 neutron-server: rpc worker ... | 08:48 |
| tkajinam | neutron 70441 70395 70441 0.0 1.7 134208 143024 neutron-server: rpc worker ... | 08:48 |
| tkajinam | I wonder if there are any other factors enabling rpc workers ? I can open a bug but would like to get some feedback in case this is an expected behavior | 08:49 |
| tkajinam | I roughly guess this is an issue caused by https://github.com/openstack/neutron/blob/master/neutron/service.py#L178 | 08:51 |
| tkajinam | hmm one might be RpcReportsWorker | 08:53 |
| opendevreview | Rodolfo Alonso proposed openstack/neutron stable/2023.1: dhcp: fix usage of helper function to retrieve process name https://review.opendev.org/c/openstack/neutron/+/907711 | 08:54 |
| *** ralonsoh_ is now known as ralonsoh | 08:56 | |
| ralonsoh | tkajinam, hi, is the rpc_worker value explicitly defined in the job? | 08:57 |
| ralonsoh | as you commented, if nothing is defined, this calculation will be done | 08:57 |
| ralonsoh | this is not a bug, RPC could be used in ML2/OVN if DHCP agents are needed | 08:57 |
| tkajinam | ralonsoh, that's the result from one of the puppet jobs and I set 0 explicitly. it seems one of these two is RpcReportsWorker but I'm unsure we still need the other RpcWorker | 08:58 |
| tkajinam | we don't have dhcp agent deployed in that scenario | 08:58 |
| ralonsoh | ah ok, you also have "rpc_state_report_workers" variable | 08:59 |
| tkajinam | #rpc_state_report_workers = 1 | 09:00 |
| tkajinam | the option is not set (it's commented out) so I assume it's effectively 1 | 09:00 |
| ralonsoh | yes, is 1 vy default | 09:01 |
| ralonsoh | by* | 09:01 |
| sahid | o/ guys do you have any idea of what is happening from a given agent in an az, when it has to handle a network with subnet on different AZs? | 09:05 |
| opendevreview | Takashi Kajinami proposed openstack/neutron master: Use different process title for RpcReportsWorker and RpcWorker https://review.opendev.org/c/openstack/neutron/+/907712 | 09:05 |
| tkajinam | ^^^ so this should change the proc title for RpcReportsWorker | 09:06 |
| sahid | s/subnet/subnets | 09:06 |
| ralonsoh | tkajinam, checking that right now | 09:06 |
| tkajinam | so my current question is... do we still need the other RpcWorker when ml2-ovn is used without any other agents such as dhcp or vpnaas ? | 09:06 |
| ralonsoh | tkajinam, no that I'm aware | 09:09 |
| ralonsoh | and if you disable it but is needed, nothing will work, so it will be easy to detect | 09:10 |
| ralonsoh | sahid, sorry, I don't understand the question | 09:10 |
| ralonsoh | subnet on different AZs? | 09:10 |
| tkajinam | ralonsoh, ah, ok. let me try removing that max function to actually pass worker 0 to RpcWorker and I'll check how it works. | 09:11 |
| opendevreview | Takashi Kajinami proposed openstack/neutron master: DNM: Allow real 0 rpc workers https://review.opendev.org/c/openstack/neutron/+/907713 | 09:16 |
| houtknots | Hello everyone, could I kindly request someone to review my recent merge request? 900212: Change exception messages to error log messages for DNS integration. | https://review.opendev.org/c/openstack/neutron/+/900212 | 09:19 |
| sahid | ralonsoh: yes it's a bit confuse, let's say you have a netowk with 2 subnets, each of the subnet are attached to physnet on different AZs | 09:20 |
| opendevreview | Takashi Kajinami proposed openstack/neutron master: Use different process title for RpcReportsWorker and RpcWorker https://review.opendev.org/c/openstack/neutron/+/907712 | 09:27 |
| opendevreview | Takashi Kajinami proposed openstack/neutron master: DNM: Allow real 0 rpc workers https://review.opendev.org/c/openstack/neutron/+/907713 | 09:27 |
| opendevreview | Takashi Kajinami proposed openstack/neutron master: Use different process title for RpcReportsWorker and RpcWorker https://review.opendev.org/c/openstack/neutron/+/907712 | 09:27 |
| opendevreview | Takashi Kajinami proposed openstack/neutron master: DNM: Allow real 0 rpc workers https://review.opendev.org/c/openstack/neutron/+/907713 | 09:27 |
| sahid | so the real question, is more like how the dhcp agent handle the case when it has to handle a network with subnets attached to different physnet | 09:27 |
| tkajinam | physnet and availability zone are both associated with networks, instead of subnets iirc or do I misunderstand something ? | 09:34 |
| tkajinam | if you deploy networks across az but your underlying networks are disconnected between az then probably something like; | 09:36 |
| tkajinam | use AZAwareWeightScheduler, deploy dhcp agents in all az, and set dhcp_agents_per_network >= (number of az) ? | 09:36 |
| tkajinam | this assumes you have ml2-ovs(or linux bridge) + dpch agent, though. I'm not very familiar with multi AZ deployment with ml2-ovn | 09:36 |
| sahid | tkajinam: yes, in that case, physnet is associated to segment and segment to subnet, it's case of routed provider network | 09:44 |
| sahid | if you have physnet per az | 09:45 |
| sahid | so you have a network with subnets on different az | 09:45 |
| opendevreview | Merged openstack/ovsdbapp master: tox: Drop envdir https://review.opendev.org/c/openstack/ovsdbapp/+/907608 | 10:44 |
| opendevreview | Takashi Kajinami proposed openstack/neutron master: Use different process title for RpcReportsWorker and RpcWorker https://review.opendev.org/c/openstack/neutron/+/907712 | 10:49 |
| opendevreview | Takashi Kajinami proposed openstack/neutron master: DNM: Allow real 0 rpc workers https://review.opendev.org/c/openstack/neutron/+/907713 | 10:50 |
| sahid | ralonsoh: ok I understans know, (in case you may be interested too), https://github.com/openstack/neutron/blob/master/neutron/api/rpc/handlers/dhcp_rpc.py#L177 | 11:05 |
| sahid | basically we return to the dhcp agent only subnets that is attached to | 11:06 |
| opendevreview | Merged openstack/os-ken master: tox: Drop envdir https://review.opendev.org/c/openstack/os-ken/+/907605 | 11:07 |
| sahid | there are so many factors :-) | 11:07 |
| opendevreview | yatin proposed openstack/neutron master: [DNM][functional] ovs-vswitchd debug log https://review.opendev.org/c/openstack/neutron/+/907037 | 11:08 |
| sahid | s/know/now | 11:12 |
| opendevreview | Rodolfo Alonso proposed openstack/neutron master: WIP - Add the external network "filter_hook" to the subnet object https://review.opendev.org/c/openstack/neutron/+/907313 | 12:13 |
| opendevreview | Merged openstack/neutron stable/2023.2: [Fullstack] Consolidate segmentation_id update tests into single test https://review.opendev.org/c/openstack/neutron/+/907226 | 12:49 |
| opendevreview | Merged openstack/neutron master: tox: Drop envdir https://review.opendev.org/c/openstack/neutron/+/907606 | 12:53 |
| opendevreview | Merged openstack/neutron stable/2023.2: dhcp: fix usage of helper function to retrieve process name https://review.opendev.org/c/openstack/neutron/+/907270 | 12:53 |
| opendevreview | Merged openstack/neutron-lib master: [sqlalchemy-20] Use sqlalchemy.orm.DeclarativeBase https://review.opendev.org/c/openstack/neutron-lib/+/906627 | 13:02 |
| sahid | there is an interesting point which does not look to be documented, in routed provider network, for a given network with subnets attached to different segment. A given agent attached to a physnet will handle only one subnet of that agent | 14:01 |
| sahid | if you configure dhcp_agents_per_network = 1 the other subnet will never be addressed by the agent on the other physnet | 14:02 |
| sahid | right? | 14:02 |
| sahid | that also means, if you want HA you need 4 DHCP Agents, 2 for each subnet/phynet/segment | 14:03 |
| opendevreview | Takashi Kajinami proposed openstack/neutron master: Use different process title for RpcReportsWorker and RpcWorker https://review.opendev.org/c/openstack/neutron/+/907712 | 15:56 |
| *** ravlew is now known as Guest1715 | 16:29 | |
| opendevreview | OpenStack Release Bot proposed openstack/networking-bagpipe master: reno: Update master for unmaintained/yoga https://review.opendev.org/c/openstack/networking-bagpipe/+/907851 | 16:45 |
| opendevreview | OpenStack Release Bot proposed openstack/networking-bgpvpn master: reno: Update master for unmaintained/yoga https://review.opendev.org/c/openstack/networking-bgpvpn/+/907853 | 16:45 |
| opendevreview | OpenStack Release Bot proposed openstack/networking-sfc master: reno: Update master for unmaintained/yoga https://review.opendev.org/c/openstack/networking-sfc/+/907856 | 16:45 |
| opendevreview | OpenStack Release Bot proposed openstack/neutron-dynamic-routing master: reno: Update master for unmaintained/yoga https://review.opendev.org/c/openstack/neutron-dynamic-routing/+/907858 | 16:45 |
| opendevreview | OpenStack Release Bot proposed openstack/neutron-lib master: reno: Update master for unmaintained/yoga https://review.opendev.org/c/openstack/neutron-lib/+/907860 | 16:45 |
| opendevreview | OpenStack Release Bot proposed openstack/neutron-vpnaas-dashboard master: reno: Update master for unmaintained/yoga https://review.opendev.org/c/openstack/neutron-vpnaas-dashboard/+/907862 | 16:46 |
| opendevreview | OpenStack Release Bot proposed openstack/neutron-vpnaas master: reno: Update master for unmaintained/yoga https://review.opendev.org/c/openstack/neutron-vpnaas/+/907864 | 16:46 |
| opendevreview | OpenStack Release Bot proposed openstack/neutron master: reno: Update master for unmaintained/yoga https://review.opendev.org/c/openstack/neutron/+/907866 | 16:46 |
| opendevreview | OpenStack Release Bot proposed openstack/os-ken master: reno: Update master for unmaintained/yoga https://review.opendev.org/c/openstack/os-ken/+/907868 | 16:46 |
| opendevreview | OpenStack Release Bot proposed openstack/ovn-octavia-provider master: reno: Update master for unmaintained/yoga https://review.opendev.org/c/openstack/ovn-octavia-provider/+/907870 | 16:47 |
| opendevreview | OpenStack Release Bot proposed openstack/ovsdbapp master: reno: Update master for unmaintained/yoga https://review.opendev.org/c/openstack/ovsdbapp/+/907872 | 16:47 |
| opendevreview | OpenStack Release Bot proposed openstack/python-neutronclient master: reno: Update master for unmaintained/yoga https://review.opendev.org/c/openstack/python-neutronclient/+/907874 | 16:47 |
| opendevreview | OpenStack Release Bot proposed openstack/tap-as-a-service master: reno: Update master for unmaintained/yoga https://review.opendev.org/c/openstack/tap-as-a-service/+/907876 | 16:47 |
| opendevreview | Merged openstack/neutron-lib master: tox: Drop envdir https://review.opendev.org/c/openstack/neutron-lib/+/907607 | 17:56 |
| JayF | Hey, we're getting some reports of Ironic+NGS+OVN users being impacted by https://review.opendev.org/c/openstack/neutron/+/872033 -- it looks like progress has stalled on it. | 18:17 |
| JayF | I'm not super knowledgable on neutron to offer help to fix, but if there's anything specific I can do to unblock it, please let me know. | 18:17 |
| mlavalle | haleyb: is Launchpad also down for you? | 18:25 |
| haleyb | mlavalle: it was up and just went back down according to my sources. i don't know the reason for the current outage :( | 18:34 |
| *** dasm is now known as Guest1739 | 20:33 | |
| haleyb | mlavalle: it's back, so hurry up and login :) | 20:37 |
| *** Guest1739 is now known as dasm | 20:38 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!