| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-09-01 | |||
| 11:34:34 | bauzas | 20 years* | |
| 11:34:52 | bauzas | now, I'm limited by my clients, not by my bandwidth | |
| 11:35:13 | sean-k-mooney | i used to have semetriic 1gbps but they apparenly only did that for an intoductray trial and dropped my upload to 200mb. i have my home cloud offline for the last while but i was actully ortinnaly more interestedin getting native ipv6 which my current isp does not have an more upload so i could host a few things here | |
| 11:35:47 | bauzas | sean-k-mooney: yeah, I'm lazy with moving to v6 | |
| 11:35:55 | bauzas | while I should | |
| 11:36:12 | bauzas | my ISP is proposing it and I could manage it | |
| 11:36:28 | bauzas | but I'm still concerned by 6to4 | |
| 11:36:34 | sean-k-mooney | i used ot use hricane eltric to provide a tunnel for ipv6 but there latency has gone up recently so removed it about 2 months ago | |
| 11:36:57 | bauzas | that's the whole problem with v6 | |
| 11:37:15 | bauzas | until the whole internet has moved, you'll face problems with 6 :( | |
| 11:37:16 | sean-k-mooney | my isp recentl up there price so im going to move before the end of the year but native ipv6 is on my list of things to look for | |
| 11:37:32 | sean-k-mooney | bauzas: oh we have dual stack connectivity | |
| 11:37:43 | sean-k-mooney | there is no ipv6 only isp in ireland | |
| 11:37:43 | bauzas | hah | |
| 11:38:10 | sean-k-mooney | most are still ipv4 only some offer native ipv6 in addtion | |
| 11:38:16 | gibi | I can write to my m2 ssd with 290MB/s you 1Gpbs download means ~128MB/s writing so a 2.5 Gpbs download would be limited by the local disk receiving the write | |
| 11:38:33 | sean-k-mooney | one even give you native ipv4 and ipv6 without CGNAT on teh ipv4 side but they are $$$$ | |
| 11:39:04 | bauzas | sean-k-mooney: we have one provider that proposes ipv6 native | |
| 11:39:10 | bauzas | but this is an option | |
| 11:39:55 | bauzas | gibi: trust me, I know it ;) | |
| 11:40:11 | bauzas | a classic dd works | |
| 11:40:40 | bauzas | but when it's about collecting pieces from the internet and reassembling them on the fly, I'm | |
| 11:40:52 | bauzas | **really IO-bound, even with 1gbps :) | |
| 11:40:57 | gibi | yeah I can imagine | |
| 11:41:35 | bauzas | yeah, reading newsgroups takes time... :) | |
| 11:42:31 | bauzas | anyway, the dudes probably have stopped to work at the network room | |
| 11:42:32 | sean-k-mooney | god i havent done that in at least a decade. i did for a short while in collage but never found it a good way to consume info | |
| 11:42:47 | bauzas | looks like my gitches are gone | |
| 11:51:15 | opendevreview | Lee Yarwood proposed openstack/nova master: workarounds: Remove disable_native_luksv1 https://review.opendev.org/c/openstack/nova/+/805647 | |
| 11:51:16 | opendevreview | Lee Yarwood proposed openstack/nova master: workarounds: Remove rbd_volume_local_attach https://review.opendev.org/c/openstack/nova/+/805648 | |
| 11:52:59 | opendevreview | Stephen Finucane proposed openstack/python-novaclient master: tests: Add missing 'nova update' tests https://review.opendev.org/c/openstack/python-novaclient/+/806916 | |
| 11:52:59 | opendevreview | Stephen Finucane proposed openstack/python-novaclient master: Add pre-commit https://review.opendev.org/c/openstack/python-novaclient/+/806915 | |
| 11:52:59 | lyarwood | bauzas / gibi ; ^ should be ready now, apologies for the mix up with the reno. | |
| 11:53:00 | opendevreview | Stephen Finucane proposed openstack/python-novaclient master: trivial: Remove use of kwargs https://review.opendev.org/c/openstack/python-novaclient/+/806918 | |
| 11:53:00 | opendevreview | Stephen Finucane proposed openstack/python-novaclient master: Microversion 2.90 - Configurable hostnames https://review.opendev.org/c/openstack/python-novaclient/+/806917 | |
| 11:53:23 | bauzas | lyarwood: ack, going lunching, but will click to look at it after | |
| 12:05:44 | gibi | lyarwood: +2 on both | |
| 12:20:52 | stephenfin | lyarwood: friendly reminder to ensure you've made any changes necessary to OSC for microversion 2.89 | |
| 12:21:28 | stephenfin | s/patch is/patch/ | |
| 13:02:40 | lyarwood | stephenfin: https://review.opendev.org/c/openstack/python-openstackclient/+/804377 | |
| 13:30:19 | gibi | sean-k-mooney: fyi just got reported from downstream https://bugs.launchpad.net/nova/+bug/1942329 | |
| 13:33:48 | sean-k-mooney | gibi: yep this is a know issue | |
| 13:34:12 | gibi | is it an older bug report open for it I should duplicate to? | |
| 13:34:17 | gibi | is there | |
| 13:34:43 | sean-k-mooney | good question tl;dr when drict physical supprot was added they did not add support for any move operations | |
| 13:35:28 | sean-k-mooney | so cold migration resize and shelve are all broken | |
| 13:35:45 | sean-k-mooney | i cant recall if i blocked it for live migration or if that is also affected | |
| 13:36:36 | sean-k-mooney | we have downstream bugs for this https://bugzilla.redhat.com/show_bug.cgi?id=1745842 | |
| 13:37:49 | sean-k-mooney | i dont see an upstream one linked so im not sure if we have an upstream bug but we have talked about it at the ptg before i think | |
| 13:38:58 | gibi | thanks | |
| 13:41:38 | sean-k-mooney | ah yes line 420 https://etherpad.opendev.org/p/nova-xena-ptg | |
| 13:42:08 | opendevreview | Merged openstack/nova stable/victoria: Reproducer unit test for bug 1860312 https://review.opendev.org/c/openstack/nova/+/802842 | |
| 13:42:15 | sean-k-mooney | no upstream links form ath either altough i did not include the downstreeam one so either but lets use your new one for now | |
| 13:44:58 | gibi | ohh cool, I have a very bad memory | |
| 13:45:54 | sean-k-mooney | this has been on my todo list to fix evenuatlly but it keep getting pushed off. artom started fixing some of the degecases and we have resolved other incidnetally. | |
| 13:46:15 | sean-k-mooney | for example live migration did not wrok because 1 the mac was not updated and 2 detach of PF was not implmented | |
| 13:46:29 | sean-k-mooney | gibi: you fixed detach when you added sriov attach/detach supprot | |
| 13:46:37 | gibi | yepp | |
| 13:46:46 | gibi | and I see artom fixed the unshelve pci_slot case | |
| 13:46:51 | sean-k-mooney | yes | |
| 13:47:18 | sean-k-mooney | so we have been trying to fix the more commens cases first when we get time too work on bugs | |
| 13:47:19 | gibi | I made TODO now on my side to look at the MAC update at some point | |
| 13:48:03 | sean-k-mooney | ack, so it a general problem for all pf move operation including evacuate | |
| 13:48:25 | sean-k-mooney | as you noted we do update the neutron port on vm boot intially its just never done after that | |
| 13:48:35 | gibi | yepp | |
| 13:48:40 | sean-k-mooney | so we need to add the same logic in all the places to update the neutron port mac | |
| 13:48:55 | sean-k-mooney | the problem is that that is not part of multipe port bidnign which makes it annoying | |
| 13:49:46 | gibi | ohh then live migration will be a special case | |
| 13:49:48 | kashyap | gibi: Is it something recent - Zuul requires uBlockOrigin to be disabled? | |
| 13:49:59 | gibi | kashyap: didn't noticed yet | |
| 13:49:59 | kashyap | (And Privacy Badger) | |
| 13:50:16 | sean-k-mooney | yep gibi so if you work on this proably leave live migration till last | |
| 13:50:24 | gibi | sean-k-mooney: ack, good idea | |
| 13:50:42 | sean-k-mooney | we could add a api block noting its not implemented in the interim with a workaround flag to allow operators to renable it | |
| 13:51:01 | sean-k-mooney | its one of the things we disucssed downstream i just never got around to pusshing a patch | |
| 13:51:28 | gibi | it is not utterly broken as we are not leaking resources or breaking the guest, so I would not block the operation | |
| 13:51:49 | sean-k-mooney | well actully it kind of is | |
| 13:52:02 | sean-k-mooney | the vm will migrate but it can cause other vms to fail to boot | |
| 13:52:09 | gibi | why? | |
| 13:52:20 | sean-k-mooney | when we release the PF on the source | |
| 13:52:29 | kashyap | When I opened a Zuul URL to debug, it was asking me to "please try with disabling ad blocker". /me tries. It's iether that, or it's just a stable URL | |
| 13:52:34 | kashyap | s/stable/stale/ | |
| 13:52:51 | sean-k-mooney | if another vm in the same network is shcudled to the source node then it will get a mac conflict in neutron when it trys to update the mac | |
| 13:53:15 | sean-k-mooney | gibi: also the vm that was migrate will not get an adress form dhcp proeprly | |
| 13:53:17 | gibi | Ohh so neutron will reject the AMC update | |
| 13:53:24 | gibi | :/ | |
| 13:53:28 | gibi | OK then it is broken | |
| 13:53:29 | sean-k-mooney | yep | |
| 13:53:31 | opendevreview | Balazs Gibizer proposed openstack/nova master: [func test] move unshelve test to the proper place https://review.opendev.org/c/openstack/nova/+/793621 | |
| 13:53:31 | opendevreview | Balazs Gibizer proposed openstack/nova master: Support interface attach / detach with new resource request format https://review.opendev.org/c/openstack/nova/+/800089 | |
| 13:53:52 | sean-k-mooney | so unless you are using static ips the vm that was migrated will be broken too | |
| 13:53:58 | gibi | kashyap: could paste a link you are strugling with? | |
| 13:54:09 | kashyap | gibi: https://zuul.opendev.org/t/openstack/build/a4b92ce40be841eaaf3e97f00c39b4d4 | |
| 13:54:17 | kashyap | gibi: It's from June, granted | |
| 13:54:21 | gibi | sean-k-mooney: ahh so the dhcp agent uses the port.mac to provide the ip | |
| 13:54:24 | kashyap | Probably it just got garbage-collected | |
| 13:54:37 | gibi | kashyap: it is already gc-d | |
| 13:54:42 | gibi | kashyap: you need to recheck the patch | |
| 13:54:45 | kashyap | Yep | |
| 13:54:53 | sean-k-mooney | gibi: yes and the port will have the mac of the new PF | |