1 |
On 30/01/2017 23:46, Grant Edwards wrote: |
2 |
> I've got a couple Gentoo machines that normally run 24/7. I've |
3 |
> learned over the years that it's a good idea to reboot them |
4 |
> occasionally (when I have some spare time and I know they're idle) |
5 |
> just to make they still can. |
6 |
> |
7 |
> I've settled on roughly once a month or so. |
8 |
> |
9 |
> What seems to happen if I don't do this is that some update (or |
10 |
> perhaps just a stupid configuration mistake on my part) will render |
11 |
> the machine non-bootable, and I won't discover it until several months |
12 |
> later at the worst possible moment when I'm in the middle of something |
13 |
> urgent and the power fails, or I type "reboot" into the wrong xterm, |
14 |
> or whatever. Or maybe those things don't happen to other people... |
15 |
> |
16 |
|
17 |
I'll wager the majority of experienced folks here do much the same as |
18 |
you, I know I do on my own boxes. |
19 |
|
20 |
One thing I've been trying to ram in at work is regular monthly reboots |
21 |
of all systems. You know how it goes - machine has 1000+ days uptime[1] |
22 |
w00t! w00t! |
23 |
|
24 |
and then the power goes off |
25 |
and then you find the drives won't spin up because the bearings are |
26 |
rumbling and the psu just can't deliver the oomph anymore to spin up all |
27 |
8 drives at once |
28 |
and then the shit really hits the fan for real! |
29 |
|
30 |
So far I can't get agreement to do it (inertia? fear of loss of street |
31 |
cred? idiotic product owners? I dunno...) |
32 |
Maybe I'll sneak a monthly repeating change control in and just do it |
33 |
|
34 |
[1] 1000 days uptime these days is stupid. All it proves is that the |
35 |
admin is not doing kernel updates and the host probably leaks security |
36 |
holes like a sieve |
37 |
|
38 |
|
39 |
-- |
40 |
Alan McKinnon |
41 |
alan.mckinnon@×××××.com |