| 1 |
On Thu, Sep 1, 2016 at 2:09 PM, Volker Armin Hemmann |
| 2 |
<volkerarmin@××××××××××.com> wrote: |
| 3 |
> |
| 4 |
> a common misconception. But not true at all. Google a bit. |
| 5 |
|
| 6 |
Feel free to enlighten us. My understanding is that data=journal |
| 7 |
means that all data gets written first to the journal. Completed |
| 8 |
writes will make it to the main filesystem after a crash, and |
| 9 |
incomplete writes will of course be rolled back, which is what you |
| 10 |
want. |
| 11 |
|
| 12 |
But simply disagreeing and saying to search Google is fairly useless, |
| 13 |
since you can find all kinds of junk on Google. You can't even |
| 14 |
guarantee that the same search terms will lead to the same results for |
| 15 |
two different people. |
| 16 |
|
| 17 |
And FWIW, this is a topic that Linus and the ext3 authors have |
| 18 |
disagreed with at points (not this specific question, but rather what |
| 19 |
the most appropriate defaults are). So, it isn't like there isn't |
| 20 |
room for disagreement on best practice, or that any two people with |
| 21 |
knowledge of the issues are guaranteed to agree. |
| 22 |
|
| 23 |
>> |
| 24 |
>> Now, I can still think of ways you can lose data in data=journal mode: |
| 25 |
>> |
| 26 |
>> * You mounted the filesystem with barrier=0 or with nobarrier; this can result |
| 27 |
> |
| 28 |
> not needed. |
| 29 |
|
| 30 |
Well, duh. He is telling people NOT to do this, because this is how |
| 31 |
you can LOSE data. |
| 32 |
|
| 33 |
|
| 34 |
>> |
| 35 |
>> * Your application didn't flush its writes to disk when it should have. |
| 36 |
> |
| 37 |
> not needed either. |
| 38 |
|
| 39 |
That very much depends on the application. If you need to ensure that |
| 40 |
transactions are in-sync with remote hosts (such as in a database) it |
| 41 |
is absolutely critical to flush writes. |
| 42 |
|
| 43 |
Applications shouldn't just flush on every write or close, because |
| 44 |
that causes needless disk thrashing. Yes, data will be lost if users |
| 45 |
have write caching enabled, and users who would prefer a slow system |
| 46 |
over one that loses more data when the power goes out should disable |
| 47 |
caching or buy a UPS. |
| 48 |
|
| 49 |
> |
| 50 |
> nope. |
| 51 |
|
| 52 |
Care to actually offer anything constructive? His advice was |
| 53 |
reasonably well-founded, even if I personally wouldn't do everything |
| 54 |
exactly as he prefers to do so. |
| 55 |
|
| 56 |
-- |
| 57 |
Rich |
Archives