Using Dashboard to Monitor Logs and View Flow

Dashboard (beta) offers the insight of running tasks in Jina. With dashboard, one can analyze logs, design flows and view Jina Hub images.

flow demo

🌟 Features:

  • Log streaming, real-time chart on log-level.

  • Grouping logs by Pods, Executors. Full text search on logs.

  • Drag & drop flow design, setting properties of each Pod via a webform.

  • Flow can be imported from/exported to YAML.

Start the Log Server

Log server is a helper thread in Jina flow. It exposes HTTP endpoints to the public which the dashboard can use to fetch logs, visualize the flow.

By default the log server is disabled. To enable it you can,

If you use Flow API in Python,
from jina.flow import Flow

f = (Flow(logserver=True)

with f:
...or write a Flow from YAML
# myflow.yml

  logserver: true
f = Flow.load_config('myflow.yml')

with f:
...or start a Flow from CLI
jina flow --logserver --yaml-path myflow.yml

Either way, if you see the following logs show up in the console, then your log server is successfully running. You can now move to the next step.

logserver success started

Connect the Dashboard to Your Log Server

Go to:

Click on the globe icon on the top-left corner to connect to the log server.

It should turn into a green check mark, which means the connection is success.

log server settings

You should now see the log-streaming and flow visualization.

If it has a red cross, it means the connection is lost or the endpoint is not set correctly. Please move to the next step for instruction.

Customize the endpoints

By default the configurations of the log server is as follows:

port: 5000
  log: /stream/log  # fetching log in SSE stream
  profile: /stream/profile  # fetching profiling log in SSE stream
  yaml: /data/yaml  # get the YAML spec of a flow
  shutdown: /action/shutdown  # shutdown the log server
  ready: /status/ready  # tell if the log server is ready, return 200 if yes

You can customize the endpoints of the log server via a YAML, say mylogserver.yml. Then pass it to the Flow API via

If you use Flow API in Python,
f = Flow(logserver=True, logserver_config='mylogserver.yml')
...or write a Flow from YAML
  logserver: true
  logserver_config: mylogserver.yml 
...or start a Flow from CLI
jina flow --logserver --logserver-config mylogserver.yml ...

Don’t forget to update endpoint in the dashboard accordingly.

log server settings

Self-host a Dashboard

One can self-host a dashboard locally.

  1. git clone && cd dashboard.

  2. Install dependencies using command yarn.

  3. Run dashboard via the following ways .

Run in the debug mode

  1. node testServer

  2. testServer will be running on http://localhost:5000 by default

  3. yarn start

  4. dashboard will be served on http://localhost:3000 by default

Run in the live mode

  1. yarn build

  2. node dashboard

  3. dashboard will be served on http://localhost:3030 by default