hgy__ | 111 | 02:24 |
---|---|---|
*** queria is now known as Guest4070 | 02:25 | |
*** queria is now known as Guest4071 | 02:30 | |
onelegend | hey | 02:35 |
*** rlandy is now known as rlandy|ruck | 10:34 | |
DK4 | Hello, i've read many docs about having separate controller and network nodes, is it feasible to have the network nodes included into the controller nodes? | 12:44 |
frickler | DK4: technically that is perfectly possible. separation is more a question for your security concept and/or scalability/performance | 12:46 |
DK4 | thanks, in my first deployment i have to go this way - scalability / performance should be fine as i have rather performant machines for the first planned amount of tenants | 12:47 |
*** Guest3656 is now known as redrobot | 13:03 | |
*** rlandy|ruck is now known as rlandy|ruck|mtg | 13:16 | |
*** sshnaidm_ is now known as sshnaidm | 14:03 | |
lowercase | Is there a way to roll images Nova is currently using to the latest build version of the glance image. For example, VM was built using an image from 2020, but currently the latest version of that image is now from 2021, if the server was ever to re rebuilt (rare to never) then nova should use the newest available image. | 14:11 |
lowercase | Victoria btdubs | 14:11 |
DK4 | but as far ive read is that the network node does the routing to external providers. if i have a setup that is spread across two countries where i have the controlplane in one region and compute / storage in another. then all compute nodes need to send their packets to the network node first to reach the internet? | 14:24 |
lowercase | yes and that sounds super slow and is an anti-pattern | 14:27 |
lowercase | Ideally you would have two AZs, one in each location. Each with thier own storage and compute nodes. Any data that needs to be replicated can now be replicated cross AZ. | 14:28 |
DK4 | so the solution would be: https://docs.openstack.org/neutron/xena/admin/deploy-ovs-ha-dvr.html | 14:28 |
lowercase | That's a whole lot of, ask someone else. In my opinion I wouldn't ever ever ever consider that to be a viable option. | 14:30 |
lowercase | my network stack is diff from yours | 14:30 |
DK4 | lowercase: you mean two AZs with each their own controlplane? | 14:31 |
lowercase | yep | 14:32 |
DK4 | yes that makes sense | 14:33 |
lowercase | It really depens on how far down you want this to go. For me, each region is there own Openstack cluster. | 14:35 |
lowercase | Splitting the haproxy config into multiple local regions wasn't really an option. We're too big and it would be too much of a hassle to configure all those AZ's with unique haproxy configs. | 14:36 |
itlinux | hi all | 15:28 |
*** rlandy|ruck|mtg is now known as rlandy|ruck | 17:33 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!