That’s an ops decision. The most common things to die are RAM and PSUs. The latter are redundant on every server I’ve ever seen. The former is dirt cheap (especially as the hardware ages), and extremely easy to replace.
I have 13 year old Dell R620s that have been running 24/7/365 for a few years in a suboptimal environment at this point (I mean, minus occasional restarts for kernel updates, brief maintenance periods, etc.). The only thing I’ve had to replace were RAM and a single PSU.
I have 13 year old Dell R620s that have been running 24/7/365 for a few years in a suboptimal environment at this point (I mean, minus occasional restarts for kernel updates, brief maintenance periods, etc.). The only thing I’ve had to replace were RAM and a single PSU.