It occurred to me yesterday that user-data can be big. EC2 is limited to 16K, but many other platforms have higher limits. That means we potentially have a very big blob of base64 encoded text in an other-wise reasonably small file.
That, coupled with the fact that for quite some time user-data has been available in /var/lib/cloud/instance/user-data.txt (badly named...) seems to mean to me that we should just not put user-data in the json.
We could still have cloud-init query make it available.
The same seems true of vendor-data.
It occurred to me yesterday that user-data can be big. EC2 is limited to 16K, but many other platforms have higher limits. That means we potentially have a very big blob of base64 encoded text in an other-wise reasonably small file.
That, coupled with the fact that for quite some time user-data has been available in /var/lib/ cloud/instance/ user-data. txt (badly named...) seems to mean to me that we should just not put user-data in the json.
We could still have cloud-init query make it available.
The same seems true of vendor-data.
And then, needs to update bash_completion also.