I’ve installed Flyte via the <flyte-binary> Single...
# ask-the-community
m
I’ve installed Flyte via the flyte-binary Single cluster simple cloud deployment path whilst testing and exploring. However, the default task resource limits are too low and I’m struggling to find where I would define these to override the defaults? Could anyone point me in the correct direction? Thanks
s
You can edit the flyte-binary-config configmap and add task resources. i.e.:
Copy code
010-inline-config.yaml: |                                                                                                                                                                                          
    plugins:                                                                                                                                                                                                         
      k8s:                                                                                                                                                                                                           
        default-env-vars:                                                                                                                                                                                            
        - FLYTE_AWS_ENDPOINT: <http://sandbox-minio.flyte:9000>                                                                                                                                                        
        - FLYTE_AWS_ACCESS_KEY_ID: minio                                                                                                                                                                             
        - FLYTE_AWS_SECRET_ACCESS_KEY: miniostorage                                                                                                                                                                  
    task_resources:                                                                                                                                                                                        
      defaults:                                                                                                                                                                                             
        cpu: 1                                                                                                                                                                                              
        memory: 4Gi                                                                                                                                                                                         
        storage: 5Gi                                                                                                                                                                                        
      limits:                                                                                                                                                                                                
        cpu: 16                                                                                                                                                                                                                                                                                                                                                                                          
        memory: 16Gi                                                                                                                                                                                         
        storage: 20Gi
Or edit the your corresponding values if you're using the helm chart.
m
Ah great, thank you - I will give that a go now
k
Ya this has been updated now
There is a PR that removes all defaults
a
This is still not working for me.
Why is there resource limits on a demo cluster?
and if so, why isn't it overridable from inside the task?
d
Hi @Abin Shahab I just reproduced this on the local sandbox (
flytectl version 0.6.45
) and the limits are applied if they land on the configmap 1.
kubectl edit cm flyte-sandbox-config -n flyte
2. Find the
100-inline-config.yaml
section and add the
task_resources
section described on this thread (adjust numbers to match your needs) That will set platform-wide resource limits, which is the max amount of resources users can request on a Task.
a
ok
I have a different question, It seems that the sandbox cluster cannot launch ray clusters
is that right?
d
unfortunately, that seems to be the case right now. There's a recent issue reported: https://github.com/flyteorg/flyte/issues/4026
a
where is the code different for the sandbox?
Where can I find the ray-dashboard?
153 Views