1 |
On Dec 25, 2012 1:55 AM, "Alan McKinnon" <alan.mckinnon@×××××.com> wrote: |
2 |
> |
3 |
> On Mon, 24 Dec 2012 06:58:15 -0600 |
4 |
> Dale <rdalek1967@×××××.com> wrote: |
5 |
> |
6 |
> The truth is simply this (derived from empirical observation): |
7 |
> |
8 |
> Long ago we had established conventions about / and /usr; mostly |
9 |
> because the few sysadmins around agreed on some things. In those days |
10 |
> there was no concept of a packager or maintainer, there was only a |
11 |
> sysadmin. This person was a lot like me - he decided and if you didn't |
12 |
> like it that was tough. So things stayed as they were for a very long |
13 |
> time. |
14 |
> |
15 |
|
16 |
The convention stuck for a loooooong time because it works, it's |
17 |
reasonable, and it does not place unduly restrictions on the SysAdmin. |
18 |
|
19 |
Even back when hard disks are a mote in the eyes of today's mammoths, you |
20 |
*can* make /usr part of /, there's no stopping you. Sure, other SysAdmins |
21 |
may scoff and/or question your sanity, but the choice is yours. YOU know |
22 |
what's best for your precious servers, YOU made the call. |
23 |
|
24 |
But with the latest udev, Lennart et al saw it fit to yank that choice out |
25 |
of the hands of SysAdmins, while at the same time trying to enforce a |
26 |
stupidly overbloated init replacement. |
27 |
|
28 |
> Thankfully, it is not like that anymore and the distinction between |
29 |
> / and /usr is now so blurry there might as well not be a distinction. |
30 |
> Which is good as the distinction wasn't exactly a good thing from day |
31 |
> 1 either - it was useful for terminal servers (only by convention) and |
32 |
> let the sysadmin keep his treasured uptime (which only proves he isn't |
33 |
> doing kernel maintenance...) |
34 |
> |
35 |
|
36 |
When you're in charge of over 100 servers as the back-end of a |
37 |
multinational company that has a revenue in excess of 10 million USD per |
38 |
day, even a temporary outage means the CIO, COO, and CEO breathing down |
39 |
your neck. |
40 |
|
41 |
There's an adage: If it ain't broke, don't fix it. |
42 |
|
43 |
> I'm sorry you bought into the crap about / and /usr that people of my |
44 |
> ilk foisted on you, but the time for that is past, and things move on. |
45 |
> If there is to be a convention, there can be only one that makes any |
46 |
> sense: |
47 |
> |
48 |
> / and /usr are essentially the same, so put your stuff anywhere you |
49 |
> want it to be. ironically this no gives you the ultimate in choice, not |
50 |
> the false one you had for years. So if your /usr is say 8G, then |
51 |
> enlarge / bu that amount, move /usr over and retain all your mount |
52 |
> points as the were. Now for the foreseeable future anything you might |
53 |
> want to hotplug at launch time stands a very good chance of working as |
54 |
> expected. |
55 |
> |
56 |
|
57 |
No. I prefer any mucking in /usr to have as small effect as possible to / |
58 |
|
59 |
That I what SysAdmins worth their salary do: compartment everything. Reduce |
60 |
interdependencies as much as possible. |
61 |
|
62 |
> The design of separate / and /usr on modern machines IS broken by |
63 |
> design. It is fragile and causes problems in the large case. This |
64 |
> doesn't mean YOUR system is broken and won't boot, it means it causes |
65 |
> unnecessary hassle in the whole ecosystem, and the fix is to change |
66 |
> behaviour and layout to something more appropriate to what we have |
67 |
> today. |
68 |
> |
69 |
|
70 |
The way I see it, it's /usr integrated into / that introduces fragility. |
71 |
Too much going on in / |
72 |
|
73 |
In case you haven't noticed, since Windows 7 (or Vista, forget which) |
74 |
Microsoft has even went the distance of splitting between C: (analogous to |
75 |
/usr) and 'System Partition' (analogous to /). The boot process is actually |
76 |
handled by the 100ish MB 'System Partition' before being handed to C:. This |
77 |
will at least give SysAdmins a fighting chance of recovering a botched |
78 |
maintenance. |
79 |
|
80 |
(Note: Said behavior will only be visible if installing onto a clean hard |
81 |
disk. If there are partitions left over from previous Windows installs, |
82 |
Win7 will not create a separate 'System Partition') |
83 |
|
84 |
So, if Microsoft saw the light, why does Red Hat sunk into darkness |
85 |
instead? |
86 |
|
87 |
> -- |
88 |
> Alan McKinnon |
89 |
> alan.mckinnon@×××××.com |
90 |
> |
91 |
> |
92 |
|
93 |
Rgds, |
94 |
-- |