Closed conor-nsurely closed 1 year ago
Not that I’m aware of, sorry. But feel free to post back anything you come up with yourself in the discussions section:)
It is by far perfect, but this is the deployment I am using. On creation, each pod will download pre-generated tiles and the valhalla config json file (this is only feasible if your area is not too large). As I understood, the valhalla container can use multiple cpu cores of a node, so I found that setting the cpu request to 0.7 makes sure only one pod runs per node.
There are most likely better ways to do this, but maybe this can be a starting point, I am also curious how this can be improved 👍
apiVersion: apps/v1
kind: Deployment
metadata:
name: valhalla-deployment
labels:
app: valhalla
spec:
replicas: 10
selector:
matchLabels:
app: valhalla
template:
metadata:
labels:
app: valhalla
spec:
volumes:
- name: shared-dir
emptyDir: {}
containers:
- name: valhalla
image: gisops/valhalla:latest
env:
- name: tile_urls
value: "url/to/custom/tiles.tar url/to/valhalla.json"
ports:
- containerPort: 8002
protocol: TCP
volumeMounts:
- name: shared-dir
mountPath: /custom_files
resources:
requests:
cpu: 0.7 # Claim 70% so no other pod can be started on the same node
Did you have a look at this project?
Did you have a look at this project?
I hadn't. This deployment was made before the operator was released. But I will see if I can get that to work with Digital Ocean. The operator might be a better starting point :)
Hey guys, just found this project.
Are there any template k8s deployments/config for Valhalla? Had a look around online but can't find anything.
Thanks!