It clearly says in the documentation that point in time backups expects number of milliseconds. But the example shows the UNIX timestamp in seconds.
Based on my own usage of the CLI, I actually think the example is correct and the documentation/naming of the option is wrong. The atlas backup restore start pointInTime does actually expect the epoch to be in seconds.
Btw, if you actually provide in milliseconds, you get this error:
POST: HTTP 400 (Error code: "INVALID_JSON_ATTRIBUTE") Detail: Received JSON for the pointInTimeUTCSeconds attribute does not match expected format. Reason: Bad Request. Params: [pointInTimeUTCSeconds]
Thanks for raising this one. I’ll just run a quick test on my own system to verify the expected measurement of time unit and then work with the team to have this updated to match.
Just to double check, can you provide the atlas cli version you’re using and the command you used to get this error: Please redact the specific project ID’s and any other sensitive information before posting here
There is a ticket now logged for this which I am monitoring and will update here when the appropriate changes are made.