You’re not doing it right. There’s absolutely a reason why enterprise linux works with a version that’s more-or-less locked in place (but for security updates, like a maintenance fork). You need to understand the value you’ve been overlooking.
ten years of a stable platform. Because, yeah, it’s not this week’s release with the glitter, but it’s also not a moving target of broken suck you have to constantly chase, and you can actually do dev.
dependencies are figured out
updates are trivial when security stuff comes out. Honestly – yum+cron is so stable and reliable, and the compatibility is part of the guarantee; so you - and your customers - don’t have to worry or - please god no - delay updates to gauge risk. Since updates are 99% of the work to avoid exploits, this goes from ‘huge risk’ to ‘no-brainer’. And you don’t have to worry that your dev environment is non-trivial to replicate on the daily.
your requirements for your software becomes ‘EL7’; maybe ‘EL7+EPEL7’ or so. Your installation process becomes ‘add repo RPM which pulls in other repo RPMs. yum install’ , and you’re already onto mere config work.
validating the install isn’t ‘did you install this ream of apps from this particular week in time, then run some wget|sh bullshit? Now run this other set of commands to confirm your installation’ – but in our case is just ‘rpm -qa|egrep’ or even an snmpwalk.
not working? Give me your one config file and your rpm-qa and we’ll replicate here trivially and find out why. (I didn’t work in support, but I liaised with them a bunch in the security work, and that was common practice) Tossing 4 lines and a <<EOF construct into a vagrant config is just so easy, now, and gives the entire machine to play with.
As someone who used to dev a notable app in the past, cross-distro problems alone made so many of the fringe OSes impossible to support, and so we didn’t. EL was the backbone because we respected what we had.
I just can’t figure out why this-week’s-glitter is more important than losing the install/support/update/validate burden by choosing a stable platform to work within. Life’s too short to support dependency hell or struggle just to replicate a failing setup in your lab for testing. Do you just not support customers?