| jingvar | EugenMayer: I use Victoria, maybe something are changed in other releases | 05:58 |
|---|---|---|
| EugenMayer | jingvar overall it seems to be practice to use fairly old openstack releases in production. e... heard rocky more then once here, also stein, both EOL / unmaintained (if those docs https://releases.openstack.org/ are correct) | 09:27 |
| EugenMayer | Vitoria is less then 6 month to probably be unmaintained | 09:27 |
| EugenMayer | Anyway the inventory might has changed, i'am not sure | 09:27 |
| jingvar | there are several things, first of all Redhat :) I think a lof of people uses RDO and used Openstack release depend on it | 14:20 |
| EugenMayer | jingvar yes seen that. I actually would have also gone with centos8 if there would not have been the sstorm in regards of stream and all the backlash with new forks | 15:02 |
| EugenMayer | jingvar but AFAIK rdo is not kolla-ansible but an own deployment automation, is it? i'am not sure since the redhat docs def. are using containers, looks like it is at least kolla based, if not just wrapping it. | 15:03 |
| EugenMayer | jingvar but AFAIK rdo was on wallaby, already having a dev version of xena on stream. The point is, RDO will no longer support centos8, stream is the new goto. I'am not sure this will be what most people decided to used centos for | 15:04 |
| jingvar | what about RHOSP | 15:34 |
| jingvar | what is an enterprise product - set of versions openstack services and support - and guaranties that it will works | 15:38 |
| EugenMayer | not sure, that is the enterprise full supported (paid) variant. TBH did not seen it yet. But the RH enterprise grades are out of my budges. I do not have the needs for the benefits so the expenses pay off (for us) | 15:38 |
| EugenMayer | Are there, currently, any efforts to make life with the container encapsulation a little easier by offering wrapper binaries or aliases like https://github.com/EugenMayer/openstack-lab/blob/master/controller/1_prepare_os.sh#L17 | 15:40 |
| EugenMayer | so one can actually run ovn-nbctl show on the host without knowing which container needs to be used for that. It is quiet a non-obvious hurdle for starters to understand which container does what | 15:40 |
| jingvar | https://docs.mirantis.com/mos/latest/release-notes.html | 15:40 |
| jingvar | as iI understand Victora | 15:41 |
| EugenMayer | never heard of Mirantis/MOS yet | 15:41 |
| jingvar | :) | 15:41 |
| EugenMayer | well it is openstack witk k8s | 15:41 |
| EugenMayer | i do not like all this bundling - way to much opinion, way to old tech | 15:42 |
| jingvar | we talked about why people use old releases | 15:43 |
| EugenMayer | e.g. for k8s we use rke2 with terraform, the ingress/storage parts are then deployed with terraform. Up and read on prod and lab with one command using vanilla stuff | 15:43 |
| EugenMayer | yeah, get it. It's the all-in-one bundle | 15:43 |
| EugenMayer | 'a stack that has been revised and certified to work' | 15:43 |
| EugenMayer | with the pros and cons - got it | 15:43 |
| EugenMayer | i guess MOS will be kops vanilla k8s | 15:45 |
| jingvar | talking about containers there one main point - Lifecycle management | 15:46 |
| jingvar | stoled from a Redhat doc | 15:49 |
| jingvar | and a little bit adopted to Kolla | 15:49 |
| jingvar | EXTERNAL_ID=$(ssh ctl01 "sudo docker exec ovn_controller ovs-vsctl get open . external_ids:ovn-remote " | awk -F: '{print $2}') | 15:50 |
| jingvar | export NBDB=tcp:${EXTERNAL_ID}:6641 | 15:50 |
| jingvar | export SBDB=tcp:${EXTERNAL_ID}:6642 | 15:50 |
| jingvar | alias ovn-sbctl="ssh ctl01 \"sudo docker exec ovn_controller ovn-sbctl --db=$SBDB $@\"" | 15:50 |
| jingvar | alias ovn-nbctl="ssh ctl01 \"sudo docker exec ovn_controller ovn-nbctl --db=$NBDB $@\"" | 15:50 |
| jingvar | alias ovn-trace="ssh ctl01 \"sudo docker exec ovn_controller ovn-trace --db=$SBDB $@\"" | 15:50 |
| jingvar | and now you can run from kayobe host simple - ovn-sbctl show (for example) | 15:51 |
| EugenMayer | yes redhat does that, but kolla does not | 16:05 |
| EugenMayer | and yes that is exactly what i do, just with an alias | 16:06 |
| EugenMayer | i do this for computes for virsh and nova (while the latter is not needed, i install novaclient on deployer and run it from there) | 16:06 |
| EugenMayer | we also developed a little helper to backup VMs with a schedule and something that automatically checks the backups and throws an notice into mattermost if something went wrong using `nova backup` | 16:07 |
| EugenMayer | something like 'backup all instances if they are not blacklisted, use a daily and weekly schedule/rotation and ensure they have been actually created". Yet fairly naive, but planned to be done in go if you continue with all this | 16:08 |
| EugenMayer | jingvar you do not need to export the NBDB - you can instead use the ovn_nb_db / ovn_sb_db | 16:11 |
| jingvar | may be | 16:12 |
| EugenMayer | it works, i'am using it all the time. Works for the sb and nb | 16:15 |
| EugenMayer | are you using OVN? Thought you are using OVS only yet. Sorry maybe i confused it | 16:20 |
| jingvar | I just installed OVN and start learining | 16:21 |
| jingvar | how looks you variant of the commnad | 16:22 |
| EugenMayer | well i need to debug the flow with ovn since out of a reason my provider_lan is not working in the lab, eventhough the setup is done via terraform and is basically 100% except the provider lan subnet. Really odd | 16:22 |
| EugenMayer | cannot share all the bin wrappers, they are in a chef cookbook, but the one i have public, those 3 are here https://github.com/EugenMayer/openstack-lab/blob/master/controller/1_prepare_os.sh#L11 | 16:23 |
| EugenMayer | so ovn_sb for sbctl / trace and ovn_nb for nbctl | 16:24 |
| EugenMayer | nova_compute for nova and nova_libvirt for virsh | 16:24 |
| jingvar | as I remmeber when HA is enabled and active sb/nb moves there need to define url to db | 16:25 |
| EugenMayer | probably, i do not have HA here so that might be the point | 16:32 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!