- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
06-17-2024 10:36 AM
Is anyone familiar with installing the Datadog agent on clusters? We're not having much luck.
We honestly might not be having the init script run since we're not seeing it in the log, but we can get just a generic "hellow world" init script to run and show in the log with no issue.
Accepted Solutions
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
07-12-2024 04:13 AM
Responding here with the solution I found. Hopefully it'll help anyone with similar issues.
First, the Datadog install script is practically a matryoshka doll- the script creates another script which creates a YAML file.
One of the consequences of that is that nothing from the original init script goes into the Databricks logs for init scripts. Instead, it writes its output to /tmp/datadog_start.log. The 2nd script that is created is also in the same directory at /tmp/start_datadog.sh, and I was able to use that to run it directly in a notebook and see the root cause of my error. The error itself was purely a syntax error, easily rectified when I was able to track it down.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
06-19-2024 12:08 PM
Is there any specific error you are receiving when running the init script? Does the run complete start up or fail due to the init script?
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
06-20-2024 06:07 AM
I'm not seeing any errors. The cluster does start up without issue.
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
07-12-2024 04:13 AM
Responding here with the solution I found. Hopefully it'll help anyone with similar issues.
First, the Datadog install script is practically a matryoshka doll- the script creates another script which creates a YAML file.
One of the consequences of that is that nothing from the original init script goes into the Databricks logs for init scripts. Instead, it writes its output to /tmp/datadog_start.log. The 2nd script that is created is also in the same directory at /tmp/start_datadog.sh, and I was able to use that to run it directly in a notebook and see the root cause of my error. The error itself was purely a syntax error, easily rectified when I was able to track it down.

