rudi3 said:
Puppet is dead. Salt is dead.
Ansible is dead at 40 minutes! I said 200 nodes with salt and it works! You are stuck in 2019!
What do you use for configuration management across 15 cheap servers?
This. Fork count and pipelining. Also set host_key_checking = False if you're rebuilding often.
pauloserver
Member OP
Ok changed to 25 and down to 12 minutes kkkkk still not 8 but better
chill infrastructure for chill people 🦫
lee_mcr
Member
12 minutes to what though? If it's just apt update and copy some configs you need to look at what your playbook is actually doing. I had a guy whose "deploy" was downloading a 2GB docker image every run.
Honey badger don't care... about downtime
I run 3 servers for my bakery website and I just ssh in and do it. Am I the only one here who thinks 15 servers sounds like a lot?
For 15 servers I would honestly consider NixOS and deploy-rs. Reproducible, atomic, no agent beyond SSH. But the learning curve is real.
NixOS is not bare metal purism, it's a cult. Give me a shell script and a cron job. I have 20 servers across three countries and I manage them with rsync and a for loop.
Hel said:
rsync and a for loop
Has anyone actually tried running ansible-pull instead of push? For 15 cheap boxes behind NAT it solves a lot of orchestration headaches.
josemcelroy
Member
Ansible-pull is underrated. I use it for a remote site in Aberdeen where the VPN is flaky. Boxes phone home when they can.