can't deploy autoscaling in a local k8s cluster
Hi, I am not sure if I should post this message here. I see some issues when practicing autoscaling using v0.24.0. The attached file contains all the steps I did and there is an error, "cgroup change of group failed" at some points. Thank you!
5 Replies
foreign-sapphire•2y ago
Hey @ido6668! Thank you for reporting this. There is actually a constellation of two issues here, one is known to us, the other one is new. The fixes should be out in a couple of days.
like-gold•2y ago
Thanks @olegvasilev
foreign-sapphire•2y ago
Hey @ido6668! Fixes are out: https://github.com/neondatabase/autoscaling/pull/811 and https://github.com/neondatabase/autoscaling/pull/803
If you don't mind me asking, how do you use autoscaling?
GitHub
runner: wait for QEMU startup and prevent panic by Omrigan · Pull R...
Previously, if QEMU failed to startup, the log output contained either:
Panic, because of reader==nil
"error":"dial unix /vm/log.sock: connect: no such file or directory","...
GitHub
neonvm-controller: Enable container-mgr by default by sharnoff · Pu...
We should have this for dev at least. We'll probably need to be careful not to prematurely enable it staging/prod, but that should be straightforward enough.
quickest-silverOP•2y ago
Thanks for the fix!
I checked out from the main branch with tag (v0.25.0) and repeated the whole process, the postgres14-disk-test- pod is running properly, and I can perform the pgbench test to observe the autoscaling up and down.
I also tried to switch the docker mode '--default-cgroupns-mode' between 'private' and 'host' on a virtual machine. so far I have not seen any issue.
basically, I want to evaluate whether this amazing vertical autoscaling can potentially help the RDS deployed on cloud.
foreign-sapphire•2y ago
Ok! Let us know of any successes or hurdles. We will be thrilled to learn about the applications by the community!