1 |
At Sun, 01 Mar 2009 13:57:30 +0200 Nikos Chantziaras <realnc@×××××.de> wrote: |
2 |
|
3 |
> Peter Humphrey wrote: |
4 |
>> As it's Sunday, here's an odd little thing. |
5 |
>> |
6 |
>> Not long ago, while booting this machine, four ext3 partitions |
7 |
>> needed checks on remount count reaching zero. They had been set to |
8 |
>> 23, 24, 25 and 26 mounts. (I didn't choose the numbers; they were |
9 |
>> allocated at the time I was creating the file system.) |
10 |
>> |
11 |
>> Now, this box does get rebooted, but hardly 23 x 24 x 25 x 26 = |
12 |
>> 358,800 times all told. At, say, two reboots per day, that would |
13 |
>> take rather a long time: a little under 500 years if my arithmetic |
14 |
>> is working. |
15 |
> |
16 |
> I think you're confused. 23 means a check each 23 mounts. With 2 |
17 |
> mounts per day, that's a check every 12 days for the first and second |
18 |
> disk. |
19 |
|
20 |
I think the point is that 23, 24, 25, 26 are relatively prime so that, |
21 |
if N is initially zero, it takes 23x24x25x26 increments initially for (N |
22 |
mod 23), ..., (N mod 26) to all again be zero. |
23 |
|
24 |
> Also, except mount count, there's also a time-based check. The check |
25 |
> happens whichever of the two expires first (otherwise, a system that |
26 |
> gets rebooted once each two months or such would get checked in a |
27 |
> timely manner.) |
28 |
|
29 |
This second point is quite valid. |
30 |
|
31 |
allan |