autoscaling
autoscaling copied to clipboard
neondatabase
Postgres vertical autoscaling in k8s
## Problem description / Motivation If we start using migrations to help with draining k8s nodes (e.g. for things like the metal node replacements part of neondatabase/cloud#6471), it will be...
In reading through the agent code, it's not particularly straight forward (unless you read the comments) which locks you need to be holding to call a specific function. There are...
## Motivation Decrease the number of up-and-running nodes when it's possible to evict VMs to other nodes.
## User stories As a follow-up to #413, we should also _generally_ make sure that if a hypothetical attacker _did_ have access to our network, they'd have a hard time...
ref https://neondb.slack.com/archives/C03TN5G758R/p1690216475962149?thread_ts=1690175203.472599&cid=C03TN5G758R This should significantly help with debugging in the event of kernel panics in the VMs — which is made much more likely by memory hotplugging. Tangentially related to...
A few things here: 1. Our config files have durations for various things, and they're largely defined as "integer number of seconds" 2. We have a bunch of ad-hoc duration...
[`k8s.io/component-base/version/version.go`](https://github.com/kubernetes/kubernetes/blob/d6fe718e19881399fbb65f693c0c0f29f8f63861/staging/src/k8s.io/component-base/version/version.go) defines the way that K8s components retrieve version information. tl;dr: there's some more variables we should set at build-time with ldflags, because the scheduler [prints them on startup](https://github.com/kubernetes/kubernetes/blob/d6fe718e19881399fbb65f693c0c0f29f8f63861/cmd/kube-scheduler/app/server.go#L154). All...
It's possible to bypass the scheduler plugin by explicitly setting the `.spec.node` field for a `Pod` — this messes up our bookkeeping, so it'd be good to have some kind...