I'd estimate each line of terraform to be an order of magnitude more risk/maintenance/faff than each line of Python.
ehhhhhhhh, i don’t know about that. different strokes and all of that, but TF describing infrequently changing resources requires next to no maintenance. i’m not sure i under what they mean by “risk” in this context though.
if you don’t know $technology well then any thing you do with it is likely to be worse/riskier than using the thing that you do know.
The need for expertise in anything beyond Python + Postgres.
this should be the first point because it explains 99% of the motivation for all of this.
Terraform providers and modules and the cloud resources evolve fairly rapidly (especially anything kubernetes-related). If you're not keeping up with the latest versions you can end up somewhat stranded when a new resource feature needs to be used and you discover unexpected tech debt from the world moving out from under you.
Yes, I think the point being made is that the infrastructure described in terraform is a much more difficult thing to change and evolve over time.
Changing a source file line and shipping a new version of the app is much easier.
They are not really comparable things at all. Even though we moved to infrastructure as code, that does not mean that the infra code is like functional application code.
Come to think of it, I guess it was a damn good lie to sell infrastructure as code as so easy it's like shipping a new app version when it's everything but..
It’s not that the TF code in itself is more complex, it’s the resources you spin up using TF that are. They are stateful, expose network interfaces, require backups, updates, cert renewals and all other kinds of operational burden.
Comments
ehhhhhhhh, i don’t know about that. different strokes and all of that, but TF describing infrequently changing resources requires next to no maintenance. i’m not sure i under what they mean by “risk” in this context though.
if you don’t know $technology well then any thing you do with it is likely to be worse/riskier than using the thing that you do know.
this should be the first point because it explains 99% of the motivation for all of this.
Terraform providers and modules and the cloud resources evolve fairly rapidly (especially anything kubernetes-related). If you're not keeping up with the latest versions you can end up somewhat stranded when a new resource feature needs to be used and you discover unexpected tech debt from the world moving out from under you.
Yes, I think the point being made is that the infrastructure described in terraform is a much more difficult thing to change and evolve over time.
Changing a source file line and shipping a new version of the app is much easier.
They are not really comparable things at all. Even though we moved to infrastructure as code, that does not mean that the infra code is like functional application code.
Come to think of it, I guess it was a damn good lie to sell infrastructure as code as so easy it's like shipping a new app version when it's everything but..
It’s not that the TF code in itself is more complex, it’s the resources you spin up using TF that are. They are stateful, expose network interfaces, require backups, updates, cert renewals and all other kinds of operational burden.