Host Uptimes (or reboot frequency best practice)

[johnha]

Hi all- I'm just curious how often you all reboot your hosts? I suppose it's the same as a more general question of how often to reboot a fairly plain debian box. But I'm just curious, anecdotally what sort of frequency you do so.

I use unattended-upgrades and it sends me an email whenever an upgrade happens, and sometimes it says reboot is recommended due to a new kernel version. But unless you have a specific hardware need or there is some specific kernel vulnerability that you could be susceptible to, I'm guessing you could have long uptimes, perhaps 6 months or more?

 

[UdoB]

Interesting question. One might differentiate between Homelab vs. Business. For me this means:

• Business: reboot as soon as a software component like the Kernel has a security problem which requires this. I might delay this a fews days only (weekend...). As I have some hundred technically skilled users with login-credentials to several machines this is a must.
• Homelab: reboot on Kernel updates. Because... why not? (unattended-updates can do that for you automatically.)
  

I can remember uptimes of several months (and years!) around the year 2000 (+/ 5 years). Since then it got shorter and shorter

Best regards

 

[ness1602]

We had some machines,usually physical which were working around 2000 days, but for VE i would recommended once in a few months, eg kernel updates.

 

[oguz]

But unless you have a specific hardware need or there is some specific kernel vulnerability that you could be susceptible to, I'm guessing you could have long uptimes, perhaps 6 months or more?

after a kernel upgrade it's recommended to reboot, in order to avoid weird issues that can come up with the mismatch.

if you're running a cluster and don't want to have downtime for your VMs, you can live-migrate them to another host (or better, use shared storage) while you reboot the other ones