| DK4 | im testing a little bit around how to deploy multi-region setups. i dont want to do some shared keystone stuff so im testing keystone federation for that use-case. would that be viable? | 07:20 |
|---|---|---|
| admin1 | gueswhat, when you create a new instance, there is a checkbox that says use volume in horizon . if you say yes, your nova boots via cinder .. | 08:56 |
| admin1 | else it goes to what is defined in the flavor and from there the hypervisor | 08:56 |
| admin1 | some "appliances" when virtualized, always need a 2nd disk by default .. so root disk = /dev/vda = where the os is installed, ephemeral disk = 2nd disk which appears as /dev/vdb | 08:58 |
| admin1 | both root and ephemeral will get deleted when the instance is deleted | 08:58 |
| admin1 | if you create a volume, that that gets saved when the instance is deleted .. it gets deleted only when you delete it from cinder | 08:59 |
| gueswhat | admin1: yes, but instances created over magnum does not support this, there is a volume driver cinder option in cluster template, but still it consume local disk on compute node | 09:17 |
| gueswhat | is possible to have two tiers in cinder? ( ssd and hdd ) how does it work? do i need to map it to flavours? | 09:22 |
| admin1 | gueswhat, you need to change config and have nova use ceph .. so that everything is via ceph and not local disks | 12:30 |
| gueswhat | admin1: do you have experience with single node ceph ? | 12:40 |
| admin1 | :) | 13:06 |
| admin1 | been deploying openstack fulltime since 2014 .. i think i have seen my fair share of config options .. | 13:07 |
| DK4 | admin1: what would be the best way to upgrade openstack versions? | 13:33 |
| admin1 | depends on the way you installed it .. kolla, redhat, ansible,salt , puppet etc .. if manual, then manually | 13:59 |
| gueswhat2 | admin1: what do you recommend for single node ceph ? is worth to replace swift with radosgw ? | 14:06 |
| admin1 | gueswhat2, it depends .. if you are testing yourself or for your own .. you can use anything that works for you and utilizes the resources/hardware you have .. if its for customers, take into account what they except from cloud ( multiple nodes on fault tolerance, 3x copies on ceph) , and future growth aspects of what will happen if you get 10X | 14:19 |
| admin1 | customers demand .. how easily you can grow etc | 14:19 |
| admin1 | so the answer is not straight black and white yes and no | 14:19 |
| admin1 | you can look into eventual consistency vs strong consistency .. to decide between swift or radosgw .. each have their usage .. but you cannot have 2 object storage .. well you can have, but you have to override the RC file with the url to use the 2nd one | 14:20 |
| gueswhat2 | i see, its a personal project, mostly i want to use mangum (6clusters, 50instances), i have single node proxmox, but storage in openstack is quit confusing,... | 14:33 |
| gueswhat2 | where is defined path for local disk in nova? | 14:34 |
| admin1 | nova.conf | 14:51 |
| gueswhat2 | state_path = /var/lib/nova ? | 14:59 |
| admin1 | instances_path | 15:07 |
| gueswhat2 | thanks, not sure hwo its working in kolla https://github.com/openstack/kolla-ansible/commit/57af5e43c40ded813ec6213c88283534f27857fa, its probably path inside nova privileged container | 15:11 |
| admin1 | kolla, ansible etc are just means to deploy .. once its deployed, its the same .. so instances_path /var/lib/nova/instances is correct | 15:19 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!