1 |
On Wed, 4 Jan 2012 18:49:29 -0500 |
2 |
Michael Mol <mikemol@×××××.com> wrote: |
3 |
|
4 |
> > As I said about my ex once, time tells. Sometimes, time is the |
5 |
> > only thing that does tell too. Reminds me of wine although I don't |
6 |
> > drink it. |
7 |
> |
8 |
> I think it's absolutely ridiculous to look at udev and mdev as winner |
9 |
> or loser. I'm not trying to be even-handed or fair in this; I just |
10 |
> think they service different needs. |
11 |
> |
12 |
> Currently, the only advantage I see for udev in a server is the |
13 |
> ability to give network interfaces meaningful names... |
14 |
|
15 |
|
16 |
Even that isn't all that useful for me. For my servers I know exactly |
17 |
which interface is which (turns out that when Dell give you 4 on-board |
18 |
nics they always come up in the same order. Pretty useful.) |
19 |
|
20 |
We do the proper thing and document every bit of hardware in a central |
21 |
repo (ocsng makes this automagic) and the way it is when the box is |
22 |
racked is the way it stays till it's switched off 5 years later. |
23 |
|
24 |
Aside from disks and RAM I've only had 2 hardware failures in 4 years |
25 |
(both were Adaptec RAID cards) so changing hardware is an unusual event |
26 |
(and rather major at that when it does happen). |
27 |
|
28 |
For me, udev is more of a hindrance in the data centre than a help. I |
29 |
simply do not need it at all, so mdev interests me a lot. |
30 |
|
31 |
On my notebooks and test/development VMs, that's different. Those need |
32 |
udev. |
33 |
|
34 |
On something as complex as a node manager, I do not believe there is |
35 |
such a thing as one-size fits all or a universal design. |
36 |
|
37 |
-- |
38 |
Alan McKinnnon |
39 |
alan.mckinnon@×××××.com |