| *** sshnaidm|off is now known as sshnaidm | 07:26 | |
| odyssey4me | hey folks - not sure who may be interested in this, but there is an ovn/bgp agent in the works... https://opendev.org/x/ovn-bgp-agent | 11:30 |
|---|---|---|
| jamesdenton_alt | i'm interested. | 11:42 |
| *** jamesdenton_alt is now known as jamesdenton | 11:42 | |
| jamesdenton | thanks odyssey4me :) | 11:42 |
| odyssey4me | awesome, good to know jamesdenton :) I'm sure the team would love feedback... they're driving this as a way of doing multi-cloud or inter-cloud connectivity | 11:55 |
| jrosser | odyssey4me: is that what we need for ipv6 prefix advertisment too? | 11:58 |
| jrosser | ipv6 is would be our biggest gap for OVN right now i think her | 11:58 |
| jrosser | e | 11:58 |
| odyssey4me | jrosser no idea, lemme try and find out where they hang out and we can put y'all in touch with them directly ... no need for me to play middle-man :) | 12:02 |
| noonedeadpunk | odyssey4me: oh, that _is_ really interesting and what we were looking for for quite a while | 12:23 |
| noonedeadpunk | as indeed I believe this is smth that should give ipv6 | 12:24 |
| noonedeadpunk | not sure how stable is that considering initial commit 1 month ago :) | 12:25 |
| noonedeadpunk | but still awesome that it has landed | 12:25 |
| odyssey4me | jrosser noonedeadpunk jamesdenton I'm told that #openstack-neutron would be a good place to ask questions about it and make contact with the team if you'd like to know more about stability, plans, etc. | 12:34 |
| noonedeadpunk | sure! thanks! | 12:35 |
| odyssey4me | spatel Not sure if OVN is something on your radar, but https://opendev.org/x/ovn-bgp-agent might be of interest if so. | 12:35 |
| mgariepy | hmm why does the doc points to 19.0.6 instead of the latest ? https://docs.openstack.org/project-deploy-guide/openstack-ansible/stein/deploymenthost.html#install-the-source-and-dependencies | 12:42 |
| spatel | odyssey4me sounds interesting, thanks i will sure take a look | 12:56 |
| spatel | https://opendev.org/x/ovn-bgp-agent/src/branch/master/doc/images/evpn_traffic_flow.png | 13:02 |
| spatel | may need some more document to understand whole process | 13:02 |
| spatel | jamesdenton do you have Intel X510T2 ? | 13:06 |
| spatel | sorry X550T2 | 13:06 |
| jamesdenton | X550? I only have X540 here, but know some environments have them | 13:34 |
| cardiff_dan | hey folks, I've got a near standard install of Openstack Wallaby on Ubuntu 18.04 and experiencing some odd behaviour: instances with an internal IP in the 10.0.3.0/24 range are failing to get internet connectivity when the iptables rule `-A POSTROUTING -s 10.0.3.0/24 ! -d 10.0.3.0/24 -j MASQUERADE` is in place on the controller nat iptable. As far as I can tell this rule is added automatically when lxc is installed | 13:39 |
| jrosser | noonedeadpunk: didnt we look at something similar to this before ^ ? | 13:40 |
| spatel | jamesdenton this is the error i am getting when i attach dpdk port - https://paste.opendev.org/show/809766/ | 14:14 |
| spatel | what driver version are you running? | 14:14 |
| jamesdenton | same driver | 14:18 |
| jrosser | cardiff_dan: it's interesting that traffic seems to be escaping the neutron router namespace and getting mixed up with stuff on the host? can you see how thats happening? | 14:19 |
| jamesdenton | spatel i had the same issue and never figured it out | 14:19 |
| spatel | really :O | 14:20 |
| spatel | even after upgrade of firmware ? | 14:20 |
| jamesdenton | this was a couple of weeks ago, and you mentioned the shared memory thing fixed yours. Didn't fix mine | 14:20 |
| spatel | official document saying, its supported | 14:20 |
| spatel | that was in HP server, now i have bunch of DELL servers i am dealing with | 14:21 |
| jamesdenton | yes, it's supported, i've done it before (with these same hosts) but it's been a couple of years | 14:21 |
| spatel | look like we have to make habit of documenting everything single thing :) | 14:21 |
| spatel | its hard to remember what we did 2 month ago.. | 14:22 |
| jamesdenton | for sure | 14:22 |
| jamesdenton | how's your OVN deploy working out> | 14:22 |
| spatel | i am planning to create blog of each server and card i have and process of DPDK / SRIOV etc.. | 14:22 |
| spatel | My OVN is running great, not a single issue | 14:22 |
| spatel | i am running DPDK / SRIOV and standard virtio all kind of workload | 14:22 |
| spatel | very stable so far.. | 14:23 |
| spatel | only issue is hard to know where and how to troubleshoot when it comes to OpenFlow related issue. (so far i didn't see any issue) | 14:23 |
| opendevreview | Andrew Bonney proposed openstack/openstack-ansible master: Remove unnecessary pki step in haproxy install https://review.opendev.org/c/openstack/openstack-ansible/+/812361 | 14:23 |
| cardiff_dan | jrosser: I think it is yeah, certainly removing the rule fixes the problem, but wasn't viable to leave that way. Can you suggest a way of identifying that's happening? | 14:24 |
| cardiff_dan | *what's | 14:24 |
| spatel | one downside i don't like about OVN is when you reload ovs switch it disturb traffic for few seconds | 14:24 |
| jamesdenton | cardiff_dan is your tenant network behind a neutron router? | 14:24 |
| cardiff_dan | jamesdenton: it is yeah | 14:28 |
| jamesdenton | is SNAT disabled on that router? | 14:28 |
| cardiff_dan | jamesdenton: SNAT is enabled | 14:33 |
| spatel | jrosser help me to understand this output - https://paste.opendev.org/show/809768/ | 15:09 |
| spatel | why VmallocTotal is so high ? on 64GB memory box | 15:09 |
| jrosser | well, vmalloctotal is the amount of virtual rather than physical memory that is allocated | 15:14 |
| spatel | that output saying 34TB am i wrong | 15:17 |
| foutatoro | Hi, I deployed openstack + ceph on 2 servers, after deployment it seems that objects storage service is up but I can't create containers nor make a "openstack container list". In my understanding when using ceph we don't need to set swift conf in openstack_user_conf.yml | 16:19 |
| foutatoro | you can find my conf and output erreor here https://paste.opendev.org/show/809771/ | 16:20 |
| spatel | foutatoro i don't think you need two storage system, ceph is all you need | 16:25 |
| foutatoro | spatel: yes I know but I'm wondering if I miss something in my conf | 16:26 |
| foutatoro | I deploy from the current master | 16:27 |
| spatel | you should look at swift error logs, may not able to authenticate keystone | 16:29 |
| spatel | I am not expert in swift so don't know.. mostly i stay away from swift since ceph covers everything | 16:30 |
| jrosser | foutatoro: you have a 503 response, so start by looking at the haproxy status | 16:31 |
| jrosser | see if the frontend/backend are up or down | 16:31 |
| foutatoro | jrosser: thanks for your suggestion haproxy call ceph-rgw containers on port 8080 for object storage services but the port 8080 is not in used into those containers | 16:43 |
| foutatoro | I have to check why ceph is not running properly | 16:44 |
| jrosser | there is a status interface to haproxy, or you can use hatop to look at the internal status of it | 17:05 |
| jrosser | if the backends are down you need to investigate the rgw containers | 17:05 |
| jamesdenton | cardiff_dan was able to sorta-replicate your issue locally. Added a second subnet, 10.0.4.0/24, behind the same router and it worked fine. 10.0.3.0/24 gets lost in the nether | 17:52 |
| cardiff_dan | jamesdenton: great, glad it isn't just me! | 18:03 |
| cardiff_dan | jamesdenton: crazier question though, any ideas how to remove the problem? | 18:06 |
| jamesdenton | not quite. i feel like we've seen something similar before, and it was related to the way linux routes packets. lxcbr0 likely owns 10.0.3.1, and when traffic is traverses the brq bridge it might hit there first rather than up into the namespace | 18:14 |
| jawad-axd | Hi folks, trying to run cloudkitty in my openstack-asnible aio environment with genocchi and ceilometer (Wallaby and Ubuntu). All telemetry services seem to be running fine, but cant see any reports on horizon. In rating summary, cant see tenants to which I did "openstack role add --user cloudkitty --project demo rating". Added hashmap service 'instance' with flavor_id (followed the docs). Some pointer would give | 18:17 |
| jawad-axd | a push on it..thanks | 18:17 |
| jawad-axd | https://paste.openstack.org/show/809777/ Journal cloudkitty. | 18:23 |
| mgariepy | so it's time to get tcpdump out @jamesdenton ? | 18:32 |
| jamesdenton | yep. this post rings a bell: https://vincent.bernat.ch/en/blog/2017-linux-bridge-isolation | 19:05 |
| mgariepy | sooo.. we should enable bridge vlan filtering by default on osa >? | 19:28 |
| jamesdenton | i don't really know | 19:42 |
| spatel | jamesdenton ty | 21:43 |
| spatel | yt | 21:43 |
| spatel | i have no idea what i did but now dpdk working with X550T | 21:43 |
| jamesdenton | you'll be happy to know i've converted my LXB to OVN. now time to fix everything | 21:43 |
| jamesdenton | hey, cool! | 21:44 |
| spatel | now trying to revert my changes to find out what i did :) | 21:44 |
| spatel | LXB to OVS ? | 21:44 |
| spatel | LXB to OVN.. oh! | 21:44 |
| spatel | did you find out how to migrate existing setup to ovn? | 21:44 |
| jamesdenton | well, sorta? i deleted all of my VMs but left everything else | 21:45 |
| jamesdenton | leaving the VMs would require some DB work that i'm just not prepared to tackle | 21:46 |
| spatel | That is awesome! | 21:47 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!
