Hi Everyone, We are facing issues with Metaflow in...
# ask-metaflow
i
Hi Everyone, We are facing issues with Metaflow installed on K8s (Metaflow Service, UI and S3 Datastore), where the workflow itself runs successfully in Argo Workflow, but not showing anywhere in the Metaflow UI (no entry on the the database
runs_v3
table as well) Installed via Helm: https://github.com/outerbounds/metaflow-tools/tree/master/k8s/helm/metaflow but with latest version tags of metaflow_service (2.3.3) and ui (1.1.4).
Copy code
{
  "METAFLOW_DEFAULT_DATASTORE": "s3",
  "METAFLOW_DATASTORE_SYSROOT_S3": "s3://****************",
  "METAFLOW_KUBERNETES_NAMESPACE": "metaflow",
  "METAFLOW_KUBERNETES_SERVICE_ACCOUNT": "metaflow-metaflow-service",
  "METAFLOW_SERVICE_URL": "<http://metaflow-metaflow-ui.metaflow.svc.cluster.local:8083/api/metadata>"
}
Thanks
s
can you also add
METAFLOW_DEFAULT_METADATA
as
service
in your config and try again?
i
With that added, I’m seeing a different error on the tasks (Tasks are failing). Still not showing up in the metaflow UI
Copy code
Code package downloaded.
Task is starting.
    Metaflow service error:
    Metadata request (/flows/AlphaFlow) failed (code 405): 405: Method Not Allowed

time="2022-12-18T19:58:30.913Z" level=info msg="no need to save parameter - on overlapping volume: /mnt/out/task_id" argo=true
Error: exit status 1
405 Method Not allowed.
The same workflow without adding that variable, successfully runs on Argo, but still not showing up in Metaflow UI
Copy code
Setting up task environment.
Downloading code package...
Code package downloaded.
Task is starting.
alpha is 0.500000
time="2022-12-18T20:01:23.820Z" level=info msg="no need to save parameter - on overlapping volume: /mnt/out/task_id" argo=true
Note: Everything works as expected if we use
2.2.4
version of the metaflow-service and
v1.0.1
for the UI works as expected. The issues happens only when using
2.3.3
and
1.1.4
May the Helm needs updated to work with new version ?
s
that variable is necessary for the metadata service to record the metadata so that the UI can correctly display the information. When you use
2.2.4
/
v1.0.1
with that variable - does everything work well?
a
I had similar issues recently and had to set my METAFLOW_KUBERNETES_NAMESPACE to "argo", maybe related
šŸ‘ 1
i
I needed that variable even with
2.2.4
/
1.0.1
@ancient-application-36103. But, it worked and showed up in the UI. You were exactly on point about the variable. My
METAFLOW_KUBERNETES_NAMESPACE
is
metaflow
and Argo is deployed in
metaflow
namespace.
a
how are you cycling between
2.2.4
and
2.3.3
version of the service?
i
Helm upgrade with new values for tags (DB is reset on each try - as this was a new install)
a
@icy-exabyte-25104 is this issue also resolved alongside the other issue?
i
No, these are 2 different installs @ancient-application-36103. The UI issue is where, we use if for all our development. This issue is a separate install where we are testing our upgrade. If this works, we will upgrade the real environment
a
Are you able to use the Metaflow Client to list flows that execute on Argo Workflows? Can you share the output of
python flow.py argo-workflows create --only-json
for the deployment where the execution is visible in the Metaflow UI and the deployment where the execution isn't visible?
i
Sorry Savin, missed your message. Will share the details asap
šŸ‘ 1
Hi Savin, Sorry for the very late reply. I actually destroyed the test cluster and have to re-create it after the holidays. Here is the JSON output from old version (working) and new version (not working)
h
@icy-exabyte-25104 did you get to the bottom of this? I am having similar issues here.