"asynchronous mount"

ACSial

ACSial

On a webpage about FreeBSD, it was stated that Linux uses an "asynchronous mount" and that this causes problems for post-crash reboots. Could someone please explain this, for the benefit of a simpleton non-geek like me...?

Also, do "hard crashes" (define?) invariably cause corruption of the OS? With Mac OS 9, I recall the computer making 'repairs' during startup, after an unintentional shutdown--is this similar?

Thanks,
Adam
 
This is outdated and imprecise information, you should ignore it. Ext3 is a journaling file system and its recovery works well. UFS2 uses soft updates, which should be just as safe and faster upon writing (though I think the GIANT remnants in the kernel eat that advantage), but recovery takes longer. It can be done in background so it's not that much of an issue.

An asynchronous mount caches data in memory and collects it to optimize for fast, serial writes. Should your system crash hard, you will not only loose data (which you also loose with UFS2 or EXT3), but you will also have an inconsistent file system table and you might be going to be unable to return your file system to a consistent state. E.g. you might have truncated files, which can cause loads of trouble as is in my opinion much worse than loosing a whole file.

A hard crash means that the system has no opportunity to do any last writing. This is the case during a power outage.

I mount /tmp asynchronously, because I don't need that file system to survive a crash.
 
Note that async is the recommended mode of mounting a ufs filesystem if you are using gjournal (precisely because there is a journal underneath, and the risk of losing data is negated. You shouldn't also need fsck).

kamikaze said:
UFS2 uses soft updates, which should be just as safe and faster upon writing (though I think the GIANT remnants in the kernel eat that advantage)
Click to expand...

If I remember correctly, UFS has been multithreaded since FreeBSD 6.
 
You must log in or register to reply here.
Back
Top