1 |
On Sun, Aug 30, 2015 at 11:54 AM, Alan McKinnon <alan.mckinnon@×××××.com> wrote: |
2 |
> Technically, we should do a power down test every 6 months or so, but |
3 |
> that turns out not to be a yes/no test in real life; it's a yes/destroy |
4 |
> test and no-one wants to make a decision either way. So we all sit in |
5 |
> limbo and wait for some exterior event to decide for us (like black-outs) |
6 |
> |
7 |
|
8 |
Half the time these are ancient services that have long been replaced |
9 |
but nobody can bring themselves to make the call to get rid of the old |
10 |
servers. Maybe there were 10M records in the database and 9.998M of |
11 |
them were migrated to a new database, but due to some issue the rest |
12 |
couldn't be, so the old server stays up just in case anybody ever |
13 |
needs the old data, and so on. |
14 |
|
15 |
Typically these would just stick around until finally some hardware |
16 |
component fails, and then it gets written off. |
17 |
|
18 |
Sadly, this course of forcing hands seems to be going away. At work |
19 |
somebody tried to hand me an ancient system to look after in my spare |
20 |
time. Apparently they just finished virtualizing it. Go figure - |
21 |
they have VAX VMs available for Linux these days. The problem is that |
22 |
KT and maintaining documentation and not being the person who gets the |
23 |
finger pointed at when something goes wrong costs the company time and |
24 |
money, and in this case for almost zero value. |
25 |
|
26 |
Usually the problems with technology aren't technical in nature... |
27 |
|
28 |
-- |
29 |
Rich |