| 1 |
On 24/04/2013 12:17, Neil Bothwick wrote: |
| 2 |
> On Wed, 24 Apr 2013 12:10:56 +0200, Alan McKinnon wrote: |
| 3 |
> |
| 4 |
>>> "Some directory operations (including unlink(2)) are not synchronous |
| 5 |
>>> on ReiserFS, which can result in data corruption with applications |
| 6 |
>>> relying heavily on file-based locks (such as mail transfer agents |
| 7 |
>>> qmail[9] and Postfix[10]) if the machine halts before it has |
| 8 |
>>> synchronized the disk." |
| 9 |
>>> |
| 10 |
>>> So I can lose stuff if the computer crashes. I don't see that as a |
| 11 |
>>> specific problem with MTAs. although they do tend to have a lot of |
| 12 |
>>> file throughput. On the other hand, I think the fact that maildir |
| 13 |
>>> uses so many files is one of the reasons I went with ResierFS in the |
| 14 |
>>> first place, running out of inodes on a mail server would not be my |
| 15 |
>>> idea of fun. |
| 16 |
>> |
| 17 |
>> I solve that problem for me in the obvious way: |
| 18 |
>> |
| 19 |
>> I pay less attention to choice of filesystem and more attention on |
| 20 |
>> rigging systems that don't crash! |
| 21 |
> |
| 22 |
> Maybe that's why I never hit this bug, I don't recall my mail server ever |
| 23 |
> crashing. |
| 24 |
> |
| 25 |
> If this mail does not arrive intact, I spoke too so^%£$£" |
| 26 |
|
| 27 |
|
| 28 |
Postfix doesn't crash (for the everyday human definition of "doesn't"). |
| 29 |
|
| 30 |
Here's my pair of MTAs: |
| 31 |
|
| 32 |
$ uptime |
| 33 |
12:24PM up 1295 days, 13:10, 1 user, load averages: 0.19, 0.20, 0.31 |
| 34 |
|
| 35 |
$ uptime |
| 36 |
12:24PM up 1925 days, 20:30, 4 users, load averages: 0.90, 0.75, 0.84 |
| 37 |
|
| 38 |
Those two just keep on accepting and dealing with mail, they do that a |
| 39 |
million times a day and according to uptime have been doing it for 10 years. |
| 40 |
|
| 41 |
|
| 42 |
-- |
| 43 |
Alan McKinnon |
| 44 |
alan.mckinnon@×××××.com |
Archives