2 comments

[ 3.2 ms ] story [ 22.2 ms ] thread
I'm actually impressed that this OS patch strategy didn't cause more outages over the years. It seems like there was no health check before starting more restarts. I'm also surprised that no one seems to have tested the patch because I would assume something similar would happen in a testing or canary environment.
"It’s risky to run all infrastructure on the exact same OS version"

simply staggering the updates and having some bake time in staging vs prod would fix this. how would maintaining two OSs be any better?