I saw in a recent talk by @sasajuric in a table that he replaces conrjobs, crontabs, scripts and more using Elixir. I have tried to use mix scripts but from my experience you always end up needing to go down and execute some OS level commands using
System.cmd which completely destroys the purpose and makes the script OS dependent.
- How to replace cronjobs with Elixir scripts? (without using anything from the OS?)
- How do you copy, replace, move, download files using just Elixir and not the OS commands we usually rely on?
I have no clue which talk you are refering to, but relying on a virtual machine that is not available everywhere is not making your stuff system independant.
If you really want your scripts to be system independant, than there is nothing you can do. Your scripts will always ever target either windows or linux or mac, you can combine linux and mac in a sense, but thats it.
Best efford system independance is done by using only things that are available in busybox and also using only simplest features of
sh, not the fancy extensions given by
zsh or others.
But to be honest, your scripts will always depend on the things you want to automate…
I’m sure his intend in the talk was not to discourage using OS level tools. It’s about simplifying the stack where it’s useful and reasonable. Not needing to install and manage redis if you have ETS can be a real work saver. Not using crontab even though it’s highly unlikely you’ll run the system on a box without crontab on the other hand doesn’t really make sense.
There seems to be some confusion here. I don’t replace cronjobs with mix scripts. What I tend to do, when it’s possible, is run periodical jobs straight from the system itself. I.e. I spawn a process which will periodically execute a job. This is not always possible (or simple), but when it is (and IME it often is), then you can consolidate more of the implementation of the system into a single place.
None of my systems uses any version of
cron. Me and my co workers have migrated to
I just used crontab as the example already mentioned. There are surely other ways to schedule stuff even in the beam itself. The point was rather that it’s a tradeoff to be made if you reimplement something in elixir (if it doesn’t exist) or if it’s simpler to use OS level tools already existing. It makes sense to move stuff into elixir if the overhead of managing the external system is extensive enough. If not just use the OS level tool until it no longer fits the above premise.