https://github.com/kyle-verhoog/datadog-python

kyle's vision of how datadog's python client should look
https://github.com/kyle-verhoog/datadog-python

Last synced: 23 days ago
JSON representation

kyle's vision of how datadog's python client should look

• Host: GitHub
• URL: https://github.com/kyle-verhoog/datadog-python
• Owner: Kyle-Verhoog
• Created: 2021-11-19T00:50:43.000Z (about 3 years ago)
• Default Branch: main
• Last Pushed: 2024-02-19T19:35:43.000Z (11 months ago)
• Last Synced: 2024-10-30T21:51:31.157Z (2 months ago)
• Language: Python
• Size: 69.3 KB
• Stars: 3
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: readme.md

Awesome Lists containing this project

README

        
# kyle's datadog python vision/proposal

_not for production use_

See [`examples/comprehensive.py`](examples/comprehensive.py) for a mostly

working example of the proposed API.

## 📈🐶 ❤️  🐍


 ✅ traces, metrics, logs, profiles, application security

 ✅ unified configuration

 ✅ trace-logs correlation by default

 ✅ trace-aware metrics

 ✅ typed and validated configuration



### API

```python

from datadog import DDClient, DDConfig

# Options are

#  - type-checked + validated

#  - available as corresponding environment vars

ddcfg = DDConfig(

        agent_url="localhost",

        datadog_site="us1.datadoghq.com",

        service="my-python-service",

        env="prod",

        version="0.01",

        tracing_enabled=True,

        tracing_patch=True,

        tracing_modules=["django", "redis", "psycopg2"],

        tracing_sampling_rules=[("my-python-service", "prod", 0.02)],

        profiling_enabled=True,

        security_enabled=True,

        runtime_metrics_enabled=True,

)

ddclient = DDClient(config=ddcfg)

# metrics

ddclient.gauge()

ddclient.measure()

ddclient.count()

ddclient.flush_metrics()

# logs

ddclient.log()

ddclient.warning()

ddclient.exception()

ddclient.info()

ddclient.debug()

log = ddclient.getLogger()

ddclient.LogHandler()  # or datadog.DDLogHandler()

ddclient.flush_logs()

# tracing

ddclient.trace()

ddclient.traced()

ddclient.patch()

ddclient.flush_traces()

# profiling

ddclient.profiling_start()

ddclient.profiling_stop()

ddclient.flush_profiles()

```

### `ddtrace-run`

I propose `datadog-run` which will install a default `DDClient`, initialized only via environment variable

to `datadog.client`. Essentially `sitecustomize.py` would just be something like:

```python

import datadog

from datadog import DDConfig, DDClient

_DEFAULT_CONFIG = dict(

  tracing_patch=True,  # different from the default when using the library manually

  # ... rest of defaults

)

datadog.client = DDClient(DDConfig(default_config=_DEFAULT_CONFIG))

```

## open questions/concerns

- What API is exposed for flushing data?

  - Unified for entire client?

    - Reuse connections/batch data for performance.

  - Must allow both automatic + manual strategies

    - Buffer size

    - Flush period

- What to use to locate an agent?

  - UDS vs HTTP(S) support

  - URL is weird/not intuitive with unix sockets

- Should config values store whether they are user defined?