| *** rm_work| is now known as _rm_work | 08:54 | |
| *** _rm_work is now known as ]rm_work | 08:55 | |
| *** ]rm_work is now known as [rm_work | 08:55 | |
| *** [rm_work is now known as zrm_work | 08:55 | |
| *** cgoncalves_ is now known as cgoncalves | 09:13 | |
| masterpe[m] | I have three loadbalancers in Openstack Ussuri, where they have multiple amphora systems, not only master and slaves but also the type "None". What is the best way the solve this. | 09:48 |
|---|---|---|
| * masterpe[m] < https://matrix.org/_matrix/media/r0/download/matrix.org/HUljsITjUkSnEKvowTAbUqkY/message.txt > | 09:48 | |
| masterpe[m] | I did try to failover for the 0a00f7fa-8d73-42c9-b484-a64aa88b73f9 loadbalancer but this only generated the "ALLOCATED" (48bed6c0-89ba-40ea-94ca-d8f8c205abb1) amphora system. | 09:49 |
| gthiemonge | masterpe[m]: perhaps you could try to failover a single amphora: openstack loadbalancer amphora failover <amp_uuid> | 11:43 |
| gthiemonge | there's also a amphora delete command but it was introduced in victoria | 11:43 |
| masterpe[m] | gthiemonge: Thank you for your response, I tryed to failover with the command "loadbalancer amphora failover" command. But this resulted in that the loadbalncer gets into the PENDING_UPDATE state and the amphora gets replaces with the "ALLOCATED | None" amphora node. | 11:48 |
| johnsom | masterpe[m]: make sure you have the current version for Ussuri and failover should work. Also, if there are amps with None, check if you have the spares pool enabled, those are likely spare amps. | 13:54 |
| spatel | johnsom morning, what is the different between nbproc and nbthread in haproxy? | 15:53 |
| spatel | currently i have only nbthread configure but not nproc | 15:54 |
| johnsom | spatel Hi. nbproc is forked processes, nbthreads is OS threads off of one process. You should never use nbproc. In fact, it has been deprecated for removal from HAProxy. | 15:55 |
| spatel | yes i have noticed in 2.4 version they have removed that | 15:56 |
| johnsom | spatel Also, if you are using 2.x version of HAProxy, you don't need nbtread unless you want to limit it or oversubscribe. It will automatically use a thread per available core. | 15:56 |
| spatel | currently i have 8 vCPU core vm and i have configure nbthread to 4 | 15:56 |
| johnsom | Yeah, it always had a large list of problems/limitations | 15:57 |
| johnsom | Ok, so it would pick 8 by default, but you are limiting it to 4 cores. | 15:57 |
| spatel | hmm | 15:57 |
| spatel | you are saying it will autodetect core and adjust nbthread? | 15:58 |
| johnsom | We have never used nbproc or nbthread in Octavia. | 15:58 |
| spatel | hmm | 15:58 |
| spatel | interesting | 15:58 |
| johnsom | Just a sec, let me find you a reference | 15:58 |
| johnsom | spatel https://www.haproxy.com/blog/haproxy-2-0-and-beyond/#cloud-native-threading-logging | 15:59 |
| spatel | we have 60k connection on our haproxy and seeing cpu usage is little high but everything working fine not seeing any issue. | 15:59 |
| johnsom | Yeah, I would not expect that to be very high usage. A single core can handle ~35k connections per second. | 16:00 |
| * johnsom generalizing. Not all cores are created equal. lol | 16:00 | |
| spatel | agreed, good to know that nbthread auto adjust based on vCPU count. i will remove that in future deployment | 16:01 |
| johnsom | Yeah, starting with 2.x | 16:01 |
| spatel | Thank you! | 16:02 |
| johnsom | Threads were not very good prior to 2.x in my opinion | 16:02 |
| spatel | yes i have seen some article where people complaining issues when using nproc/threads | 16:03 |
| spatel | in earlier version | 16:04 |
| opendevreview | Michael Johnson proposed openstack/octavia-dashboard master: Change the Octaiva Barbican namespace https://review.opendev.org/c/openstack/octavia-dashboard/+/800310 | 16:51 |
| johnsom | gthiemonge ^^^ renamed those files and re-tested. | 16:51 |
| haleyb | Octaiva, is that a new project? :-p | 16:53 |
| johnsom | blah | 16:53 |
| johnsom | Yes, new, improved, shiny, better than kubernetes, slices, dices, Juliannes | 16:54 |
| opendevreview | Michael Johnson proposed openstack/octavia-dashboard master: Change the Octavia Barbican namespace https://review.opendev.org/c/openstack/octavia-dashboard/+/800310 | 16:55 |
| johnsom | No good dead of hammering out a fix on Friday goes un-punished | 16:55 |
| haleyb | you're just full of typos today | 16:55 |
| johnsom | Probably | 16:55 |
| johnsom | sigh, yeah | 16:55 |
| gmann | johnsom: gthiemonge we need to cut the eol tag for stable/stein or older EM stable branches before we retire neutron-laas completely https://review.opendev.org/c/openstack/project-config/+/800147 | 17:54 |
| gmann | johnsom: and can you check this https://review.opendev.org/c/openstack/octavia/+/799405 | 17:55 |
| johnsom | gmann Sure | 17:56 |
| gmann | thanks | 17:56 |
| opendevreview | Ghanshyam proposed openstack/octavia master: Fix oslo policy DeprecatedRule warnings https://review.opendev.org/c/openstack/octavia/+/799405 | 18:10 |
| opendevreview | mitya-eremeev-2 proposed openstack/octavia master: More logging during load balancer creation. https://review.opendev.org/c/openstack/octavia/+/792697 | 19:55 |
| johnsom | rm_work You might have an opinion on these logging changes too: https://review.opendev.org/c/openstack/octavia/+/792697 | 21:16 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!