| *** tosky has quit IRC | 00:14 | |
| *** spatel has joined #openstack-ansible | 01:19 | |
| *** spatel has quit IRC | 01:23 | |
| *** gyee has quit IRC | 01:38 | |
| *** macz_ has quit IRC | 02:08 | |
| *** cshen has joined #openstack-ansible | 02:30 | |
| *** cshen has quit IRC | 02:34 | |
| *** macz_ has joined #openstack-ansible | 03:20 | |
| *** d34dh0r53 has quit IRC | 03:24 | |
| *** macz_ has quit IRC | 03:25 | |
| *** d34dh0r53 has joined #openstack-ansible | 03:27 | |
| *** cshen has joined #openstack-ansible | 04:30 | |
| *** cshen has quit IRC | 04:35 | |
| *** ianychoi has joined #openstack-ansible | 04:36 | |
| *** rh-jlabarre has quit IRC | 04:58 | |
| *** macz_ has joined #openstack-ansible | 05:02 | |
| *** macz_ has quit IRC | 05:06 | |
| *** evrardjp has quit IRC | 05:33 | |
| *** evrardjp has joined #openstack-ansible | 05:33 | |
| *** kleini has quit IRC | 05:35 | |
| *** macz_ has joined #openstack-ansible | 06:29 | |
| *** cshen has joined #openstack-ansible | 06:31 | |
| *** macz_ has quit IRC | 06:34 | |
| *** cshen has quit IRC | 06:35 | |
| *** cshen has joined #openstack-ansible | 06:37 | |
| *** cshen has quit IRC | 06:42 | |
| *** pcaruana has joined #openstack-ansible | 07:49 | |
| *** rpittau|afk is now known as rpittau | 08:03 | |
| *** cshen has joined #openstack-ansible | 08:10 | |
| *** andrewbonney has joined #openstack-ansible | 08:10 | |
| *** miloa has joined #openstack-ansible | 08:22 | |
| *** tosky has joined #openstack-ansible | 08:44 | |
| *** mmethot_ has quit IRC | 09:14 | |
| *** mmethot_ has joined #openstack-ansible | 09:14 | |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/ansible-role-systemd_service master: Add possibility to configure systemd sockets https://review.opendev.org/763194 | 09:16 |
|---|---|---|
| *** luksky has joined #openstack-ansible | 10:18 | |
| *** d34dh0r53 has quit IRC | 10:21 | |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/openstack-ansible-os_neutron stable/ussuri: Add neutron_keepalived_no_track variable https://review.opendev.org/763336 | 10:22 |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/openstack-ansible-os_neutron stable/train: Add neutron_keepalived_no_track variable https://review.opendev.org/763337 | 10:22 |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/openstack-ansible-os_neutron stable/train: Add neutron_keepalived_no_track variable https://review.opendev.org/763337 | 10:29 |
| openstackgerrit | Dmitriy Rabotyagov (noonedeadpunk) proposed openstack/openstack-ansible-os_neutron stable/train: Add neutron_keepalived_no_track variable https://review.opendev.org/763337 | 10:29 |
| *** kleini has joined #openstack-ansible | 11:06 | |
| *** pto has quit IRC | 11:31 | |
| *** pto_ has joined #openstack-ansible | 11:31 | |
| *** pto_ has quit IRC | 11:44 | |
| *** pto has joined #openstack-ansible | 11:45 | |
| *** macz_ has joined #openstack-ansible | 11:46 | |
| *** macz_ has quit IRC | 11:51 | |
| *** pto has quit IRC | 12:35 | |
| *** pto has joined #openstack-ansible | 12:36 | |
| *** gshippey has joined #openstack-ansible | 12:40 | |
| *** rh-jlabarre has joined #openstack-ansible | 13:07 | |
| *** jbadiapa has joined #openstack-ansible | 13:09 | |
| *** rpittau is now known as rpittau|brb | 13:16 | |
| *** watersj has joined #openstack-ansible | 13:46 | |
| *** spatel has joined #openstack-ansible | 13:59 | |
| *** cshen has quit IRC | 14:10 | |
| fanfi | hi all, could someone help me please, how I can enable Image-Volume cache for ceph backend ? image_volume_cache_enabled = True | 14:12 |
| *** pto has quit IRC | 14:14 | |
| noonedeadpunk | I think you should just set `image_volume_cache_enabled: true` while defining backends for cinder | 14:14 |
| *** pto has joined #openstack-ansible | 14:14 | |
| fanfi | oki, it will be easy than :) i will try this way | 14:16 |
| fanfi | thx | 14:16 |
| noonedeadpunk | oh, I missed the part where you need extra tenant for that | 14:16 |
| spatel | jrosser: noonedeadpunk can you check this - http://paste.openstack.org/show/800212/ | 14:17 |
| spatel | look like i need flat network block back for octavia right? or just add flat mechanism in neutron server? | 14:18 |
| *** pto has quit IRC | 14:19 | |
| spatel | I meant just add flat in this place on neutron server | 14:19 |
| -spatel- tenant_network_types = vxlan,vlan,flat | 14:19 | |
| -spatel- type_drivers = flat,vlan,vxlan,local | 14:19 | |
| noonedeadpunk | fanfi: well we create only cinder tenant which is used for service. ofc you can use it or create manually new one without admin privileges. after this you will be able to set `cinder_internal_tenant_project_id` and `cinder_internal_tenant_user_id` in cinder.conf with cinder_cinder_conf_overrides | 14:20 |
| noonedeadpunk | spatel: I think we do have assumption that flat network is present and tbh I never tried doing otherwise | 14:22 |
| spatel | noonedeadpunk: i thought i am using all vlan so i have removed flat type and now realized it need for octavia | 14:23 |
| noonedeadpunk | well, maybe it's not really needed - not sure tbh.... | 14:23 |
| spatel | Octavia use lbaas-mgmt on flat network.. look like | 14:24 |
| spatel | based on error when its trying to create mgmt network and saying i don't have provider | 14:24 |
| noonedeadpunk | I have no idea, honestly.. Maybe it's possible to make it on vlan network instead... | 14:25 |
| *** d34dh0r53 has joined #openstack-ansible | 14:25 | |
| spatel | http://paste.openstack.org/show/800213/ | 14:25 |
| spatel | let me add flat type in neutron and see if it make sense | 14:26 |
| noonedeadpunk | eventually you will need mapping then as well which is equal to fully returning of the section back | 14:27 |
| spatel | block in openstack_user_config.yml ? | 14:27 |
| noonedeadpunk | yeah, as you need to tell neutron what flat network is on host/container | 14:28 |
| spatel | hmm | 14:31 |
| gixx | Hello. Is it possible/viable to jump multiple releases at once? I.e. not upgrading incrementally until desired release. Running Queens at the moment, planning to upgrade to Ussuri and it'd be nice to not have to OS reinstall twice | 14:35 |
| noonedeadpunk | gixx: well, I was running direct rocky -> train upgrades, but I'd say it's something not tested, as Q->U really _a lot_ of things has changed and in upgrade script we are storing only relevant upgrade bits | 14:36 |
| noonedeadpunk | ie galera upgrade might be tricky for instance | 14:37 |
| noonedeadpunk | so you should be really aware of what you're doing | 14:37 |
| gixx | We're fine with doing some manual things to get it working | 14:37 |
| noonedeadpunk | but techincally I think it's feasable. | 14:38 |
| gixx | DB upgrades for each project should be fine? As they build upon each other, all intermediate ones should be performed in order as I've interpreted it | 14:39 |
| *** yolanda has quit IRC | 14:40 | |
| noonedeadpunk | yeah, you're right, DB upgrades of project should be done properly, as they will try to perform all chain of migrations, so this should be ok | 14:40 |
| noonedeadpunk | the thing that you should take care of is placement separation which is done in S->T and galera upgrade itself | 14:41 |
| noonedeadpunk | as there would be pretty solid jump -we use 10.4 in U | 14:41 |
| gixx | Alright thanks for the heads up | 14:41 |
| noonedeadpunk | and ofc you should check all of your overrides | 14:42 |
| noonedeadpunk | and probably add new as behaviour of services might change between releases | 14:42 |
| gixx | Yeah, I'm sure we've done noooothing custom at all there.. | 14:43 |
| gixx | Hint hint custom cloudkitty and adjutant projects :) Will be nice to get rid of the custom stuff | 14:43 |
| noonedeadpunk | well, we have cloudkitty role for a while and about to add adjutant :) | 14:44 |
| noonedeadpunk | hopefully for V | 14:44 |
| gixx | Did you get our role working eventually in master? Or did you have to scrap it all and start anew? | 14:44 |
| gixx | To be honest I kind of forgot all about it, sorry | 14:44 |
| noonedeadpunk | well, I did that patch:) https://review.opendev.org/#/c/756313/ | 14:45 |
| noonedeadpunk | seems everybody is scared to approve it lol | 14:45 |
| gixx | lol | 14:45 |
| noonedeadpunk | but it was passing and seemed working | 14:45 |
| gixx | I'm not proud if it at all. Getting py3 working in a Q deployment is not pretty at all | 14:46 |
| noonedeadpunk | adjutant has not tempest, so.... | 14:46 |
| gixx | So I'm sure it weren't up to standards at all | 14:46 |
| noonedeadpunk | eventually we're just using uwsgi where we can, and things really changed dramatically overall | 14:46 |
| noonedeadpunk | like having systemd_role and same across roles and synced db_setup and service_setup files | 14:47 |
| noonedeadpunk | It still needs some work, but I hope it can be done on top of this patch... | 14:48 |
| gixx | That's nice, makes things much simpler and cleaner | 14:48 |
| noonedeadpunk | I didn't have much time to work on it either | 14:49 |
| noonedeadpunk | so yeah:( | 14:49 |
| *** rpittau|brb is now known as rpittau | 14:51 | |
| noonedeadpunk | btw cores, can we really merge this?:) https://review.opendev.org/#/c/756313/ | 14:52 |
| noonedeadpunk | and iterate on it futher? | 14:53 |
| noonedeadpunk | as we can't do proper integrated testing until we got it merged (and can't set https://review.opendev.org/#/c/756310/ dependant on it either) | 14:54 |
| gixx | So in short, there's some database changes in S->T we have to take into consideration and manually upgrade galera. Check all overrides as well as new defaults. The S->T placement separation, where can I find more information regarding it? Is it this (https://docs.openstack.org/placement/latest/admin/upgrade-to-stein.html)? So we have to upgrade nova to Stein release, perform the database split and then resume | 15:12 |
| gixx | with rest of the upgrade | 15:12 |
| *** tosky has quit IRC | 15:18 | |
| *** tosky has joined #openstack-ansible | 15:22 | |
| *** tosky has quit IRC | 15:26 | |
| *** macz_ has joined #openstack-ansible | 15:39 | |
| noonedeadpunk | gixx: eventually in S and T we used same galera version but from Q to U it changed from 10.1 to 10.4. Regarding placement upgrade - yes, that's the right link. We had https://review.opendev.org/#/c/664867/ and https://review.opendev.org/#/c/689145/ (but not limited to) | 15:46 |
| noonedeadpunk | the thing is that we've dropped https://opendev.org/openstack/openstack-ansible-os_placement/src/branch/stable/train/templates/placement-migrate-db.rc.j2 in U | 15:46 |
| spatel | noonedeadpunk: i found my octavia issue related flat networking, we don't need flat networking. This part of missing - http://paste.openstack.org/show/800222/ | 15:47 |
| *** tosky has joined #openstack-ansible | 15:50 | |
| *** miloa has quit IRC | 15:52 | |
| noonedeadpunk | yeah, cool, as I thought you can choose network to be used:) | 15:53 |
| spatel | anyway all good, but that error message was little misleading. it should say flat provider is disabled | 15:59 |
| spatel | shouldn't* | 15:59 |
| spatel | jamesdenton: are you around? had question related Trex packet generator | 16:00 |
| jamesdenton | i'm around now | 16:04 |
| *** miloa has joined #openstack-ansible | 16:12 | |
| spatel | I have Gen9 blades which has only two NIC card and i believe for Trex we need min 3 nic card, 1 for ssh and other two for DPDK RX and TX right? | 16:14 |
| *** evrardjp has quit IRC | 16:17 | |
| spatel | jamesdenton: ^^ | 16:19 |
| jamesdenton | good question. | 16:23 |
| jamesdenton | i would think so. but maybe iLo is OK? | 16:24 |
| spatel | that is what i am thinking to use iLO for mgmt and use both nic for traffic | 16:31 |
| *** miloa has quit IRC | 16:32 | |
| *** mgariepy has quit IRC | 16:34 | |
| *** gyee has joined #openstack-ansible | 16:34 | |
| *** mgariepy has joined #openstack-ansible | 16:36 | |
| jrosser | spatel: sorry missed the octavia question - it does not need to be a flat network if you want to use vlan | 16:43 |
| jrosser | personally i would use vlan always then there is no need to mess with the config on the compute nodes to define the flat network | 16:44 |
| *** gyee has quit IRC | 16:50 | |
| *** cshen has joined #openstack-ansible | 16:51 | |
| *** rpittau is now known as rpittau|afk | 16:52 | |
| *** cshen has quit IRC | 16:55 | |
| spatel | jrosser: totally it was my fault that i missed to define octavia parameter. | 16:59 |
| *** d34dh0r53 has quit IRC | 17:02 | |
| jamesdenton | spatel i recently deployed octavia w/o the br-lbaas bridge, just configured a lbaas-mgmt network as a vlan network (like a provider network) and made sure it was reachable from control plane net. worked OK | 17:03 |
| *** macz_ has quit IRC | 17:05 | |
| *** gyee has joined #openstack-ansible | 17:05 | |
| *** d34dh0r53 has joined #openstack-ansible | 17:06 | |
| *** bverschueren has quit IRC | 17:06 | |
| *** bverschueren has joined #openstack-ansible | 17:08 | |
| spatel | jrosser: jamesdenton this method i am using to deploy Octavia with br-lbaas (james - do you using different way?) - https://satishdotpatel.github.io//openstack-ansible-octavia/ | 17:08 |
| jamesdenton | spatel http://paste.openstack.org/show/799855/ https://ibb.co/K5NcCjZ | 17:11 |
| jamesdenton | alright, i have to step out. bbiaf | 17:12 |
| spatel | jrosser: oh wait can you show me your openstack_user_config.yml ? | 17:14 |
| spatel | This is what i have http://paste.openstack.org/show/800227/ | 17:15 |
| spatel | you are telling me i don't need br-lbaas block in openstack_user_config.yml? | 17:15 |
| jrosser | no, you have to have it there in order to assign IPs to eth14 on the ocavia containers if you do it like you are currently | 17:17 |
| *** bverschueren has quit IRC | 17:18 | |
| *** bverschueren has joined #openstack-ansible | 17:18 | |
| *** klamath_atx has joined #openstack-ansible | 17:20 | |
| jrosser | spatel: james has shown you a config where you could get rid of eth14 entirely on the octavia containers | 17:20 |
| jrosser | you can choose which way you want to do it | 17:20 |
| spatel | i am ok to remove br-lbaas block if that make sense and simplify deployment. ( trying to understand if i remove br-lbaas then how octavia container wire up with VLAN 62 in my case to go out) | 17:23 |
| *** pcaruana has quit IRC | 17:25 | |
| spatel | what is octavia_service_net_setup: True does here? | 17:25 |
| jrosser | what james has done is use the default route on the container to directly get to the lbaas network | 17:26 |
| jrosser | so you have to configure your hardware router (hopefully firewall) to allow that traffic | 17:26 |
| jrosser | so 'it depends' if thats suitable or not | 17:26 |
| spatel | jrosser: oh! i know what you saying. | 17:27 |
| spatel | i have full control on my router and i can make it routable but let me stick to my way because i have same method in other datacenter so it will make documentation easy | 17:28 |
| *** pcaruana has joined #openstack-ansible | 17:29 | |
| *** pcaruana has quit IRC | 17:31 | |
| *** pcaruana has joined #openstack-ansible | 17:32 | |
| *** mmercer has quit IRC | 17:35 | |
| *** macz_ has joined #openstack-ansible | 17:53 | |
| *** pcaruana has quit IRC | 17:54 | |
| *** macz_ has quit IRC | 18:08 | |
| *** gshippey has quit IRC | 18:14 | |
| ThiagoCMC | Hey guys, I'm wondering here about the minimum requirements for using letsencrypt with OSA. So far, I found two pages: https://docs.openstack.org/openstack-ansible-haproxy_server/ussuri/configure-haproxy.html and https://docs.openstack.org/openstack-ansible/ussuri/user/security/ssl-certificates.html | 18:34 |
| admin0 | ThiagoCMC, provide the cert and it will work | 18:35 |
| admin0 | private ip -> internal IP mapping can happen in haproxy , vyos router or physical router, so there is no fixed place where its done | 18:35 |
| ThiagoCMC | Can I just stick with the "haproxy_ssl_letsencrypt_*" vars and ignore the rest? | 18:35 |
| ThiagoCMC | Sure... | 18:36 |
| admin0 | i just generate it from zerossl ( gui based) and then provide to the right location | 18:36 |
| admin0 | as most of the time, its internal or 1:1 is provided by router | 18:36 |
| ThiagoCMC | Yeah, I have self-signed now but, some services complain about it, like Magnum and a few others | 18:36 |
| admin0 | use zerossl :) | 18:36 |
| admin0 | use a real cert | 18:36 |
| ThiagoCMC | Hmm... LOL | 18:37 |
| ThiagoCMC | Checking it now! | 18:37 |
| ThiagoCMC | :-D | 18:37 |
| admin0 | its the same free cert as letsencrypt .. just that you can use the GUI to generate it | 18:37 |
| admin0 | so no need to install certbot etc in osa servers | 18:37 |
| ThiagoCMC | Nice! Letsencrypt is so cool! | 18:37 |
| ThiagoCMC | BTW, the command `openstack coe cluster list --insecure` returns: "SSL exception connecting to https://10 (HTTP 500)" - I bet that's because of my creepy self-signed cert, right? | 18:38 |
| ThiagoCMC | Masakari also doesn't like it | 18:38 |
| admin0 | having a real(free) ssl should be a pre-deployment requirement | 18:39 |
| ThiagoCMC | Very true. | 18:39 |
| *** NewJorg has quit IRC | 18:42 | |
| *** andrewbonney has quit IRC | 18:45 | |
| *** NewJorg has joined #openstack-ansible | 18:49 | |
| *** cshen has joined #openstack-ansible | 18:51 | |
| *** macz_ has joined #openstack-ansible | 18:51 | |
| *** cshen has quit IRC | 18:56 | |
| openstackgerrit | Merged openstack/openstack-ansible master: Added Netplan examples for a Openstack Infrastructure node. https://review.opendev.org/760090 | 19:04 |
| ThiagoCMC | When I click at the above "gitweb" link, it shows "Not found". Link: https://review.opendev.org/gitweb?p=openstack%2Fopenstack-ansible.git;a=commitdiff;h=461071f1baedad7851eecb78428e750cd752d4d9 | 19:14 |
| ThiagoCMC | Just me? | 19:14 |
| *** macz_ has quit IRC | 19:19 | |
| *** macz_ has joined #openstack-ansible | 19:37 | |
| *** cshen has joined #openstack-ansible | 19:55 | |
| *** watersj has quit IRC | 20:02 | |
| *** watersj has joined #openstack-ansible | 20:04 | |
| spatel | do you guys use MTU 9000 for everything or just ceph storage network? | 20:21 |
| *** luksky has quit IRC | 20:26 | |
| *** luksky has joined #openstack-ansible | 20:26 | |
| admin0 | everything possible :) | 20:36 |
| admin0 | br-mgmt and br-vxlan also | 20:36 |
| ThiagoCMC | spatel, I just realized how slow is the RBD access from my compute nodes... Gonna try 9000 here! =P - I might need a 2.5Gbps switch lol | 20:38 |
| ThiagoCMC | I'm wondering if there are multiqueue to access RBD block devices, or something else. | 20:39 |
| spatel | how much slow your RBD access? | 20:39 |
| spatel | run iperf test and see | 20:40 |
| spatel | I would definitely say use 9000 MTU for br-storage | 20:40 |
| spatel | admin0: i like to move to MTU 9000 but in my case sometime firewall other device comes in path and if they have MTU 1500 then it will be more work from them to break packets | 20:41 |
| ThiagoCMC | I'm working to get the numbers but, so far, from an instance with rbd-based virtual disc, the bonnie++ doesn't even work properly, kernel prints timeout and it's huge iowait. I missed something, for sure. | 20:51 |
| ThiagoCMC | Maybe some image properties... lol | 20:52 |
| *** jbadiapa has quit IRC | 20:52 | |
| mgariepy | ThiagoCMC, use fio. | 20:57 |
| ThiagoCMC | mgariepy, I just saw some messages at ceph mail list telling to use fio! :-D | 20:58 |
| ThiagoCMC | On it! | 20:58 |
| ThiagoCMC | Thanks! | 20:58 |
| mgariepy | haha ;D | 20:58 |
| spatel | what kind of SSD you have, that is very important because i used some cheap consumer SSD and found that was the bottleneck | 21:03 |
| spatel | later i replaced all those SSD with enterprise SSD and boom!! | 21:03 |
| ThiagoCMC | Hmm... I have Samsung SSDs, for Ceph DBs | 21:04 |
| spatel | what model? | 21:04 |
| spatel | I do have Samsung also | 21:04 |
| spatel | if its EVO or PRO then don't expect any performance | 21:04 |
| ThiagoCMC | EVO 860 | 21:05 |
| ThiagoCMC | hehe | 21:05 |
| spatel | good luck :) | 21:06 |
| ThiagoCMC | Well, I can "rbd ls" from the compute node bare-metal, I'm going to map the volumes and run some tests outside of the virtual instance to see. | 21:06 |
| spatel | you need DCT Series | 21:06 |
| ThiagoCMC | I see | 21:06 |
| ThiagoCMC | Thanks for the tip! | 21:06 |
| ThiagoCMC | It's waaaay too slow from the QEMU Instance. | 21:06 |
| ThiagoCMC | I'll see what I can get from the compute node itself. | 21:07 |
| spatel | first make sure you don't have network bottleneck (and then try new SSD) | 21:07 |
| ThiagoCMC | Even the "apt install fio" within the QEMU instance is taking a long time | 21:07 |
| ThiagoCMC | ok | 21:07 |
| spatel | We have hired Ceph consultant to fix our issue and they said use Samsumg PM863a SSDs | 21:09 |
| ThiagoCMC | That sounds expensive | 21:09 |
| ThiagoCMC | My cloud is at my basement lol | 21:10 |
| spatel | then you are good with EVO :) | 21:10 |
| spatel | i thought you are running some enterprise cloud.. | 21:10 |
| spatel | in that case EVO should be good man!! | 21:10 |
| ThiagoCMC | I was, at me previous job. But now I just want to build my own | 21:10 |
| ThiagoCMC | Sure! I believe that I should get faster access with what I have | 21:11 |
| ThiagoCMC | I'll compare the rbd from the compute node itself, with virtualization libvirt<->rbd<->virtio-scsi in the mix. | 21:11 |
| spatel | when you saying slow (what is the definition of slow ) | 21:12 |
| spatel | your VM disk located on ceph right? | 21:12 |
| ThiagoCMC | yep | 21:13 |
| ThiagoCMC | Like, 99 iowait when running "apt install fio", and it took a long time to finish | 21:13 |
| spatel | do you have (glance image-update --property hw_scsi_model=virtio-scsi --property hw_disk_bus=scsi ) ? | 21:13 |
| spatel | do you have 1G network? | 21:14 |
| ThiagoCMC | Yep, Gigalan network, and I didn't changed the image properties! | 21:15 |
| ThiagoCMC | Gonna try it now. | 21:15 |
| admin0 | ThiagoCMC, "fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --filename=test --bs=4k --iodepth=64 --size=4G --readwrite=randrw --rwmixread=75" | 21:17 |
| ThiagoCMC | admin0, awesome!!! Thank you! | 21:17 |
| admin0 | first ran on atl when testing that qnap over nfs and iscsi | 21:18 |
| ThiagoCMC | Oh, damn... I remember that! lol | 21:18 |
| admin0 | i had all the fio commands and their output saved in onenote .. | 21:18 |
| admin0 | TBW for samsung is not that much .. 1200TB | 21:27 |
| admin0 | compared to intel ones, with 10200 TB | 21:28 |
| ThiagoCMC | Damn | 21:28 |
| ThiagoCMC | admin0, have you used the fio against the block device itself? I want to map Ceph's volume at the compute node, using "rbd map" thing and benchmark agains, let's say "/dev/rbd0" | 21:31 |
| admin0 | i have not .. i did it for the instance ephemeral disk ( via ceph) and cinder ( via nfs, and iscsi) | 21:32 |
| ThiagoCMC | ok | 21:32 |
| admin0 | unless the stakeholders say lets benchmark /dev/rbd0 from kvm directly and decide if we use openstack or not, i don't see a value in getting that benchmark .. we just end up with 2 numbers :) | 21:33 |
| admin0 | your time is better used by getting magnum, octavia, trove, ovn to work :D | 21:34 |
| ThiagoCMC | LOL | 21:34 |
| ThiagoCMC | I'll get ther! | 21:34 |
| ThiagoCMC | Maybe "fio --filename=/dev/rbd0 ... " | 21:34 |
| ThiagoCMC | What about that Tacker thing, are you guys using it? | 21:35 |
| admin0 | nope | 21:36 |
| *** luksky has quit IRC | 21:41 | |
| *** watersj has quit IRC | 21:47 | |
| *** luksky has joined #openstack-ansible | 21:54 | |
| *** spatel has quit IRC | 22:16 | |
| *** cshen has quit IRC | 22:42 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!