When I moved up from Linux 2.1.xxx/RH 5.2 to now 2.3.11/ RH 6.0, I
noticed something odd:
Now when my test platform crashes (due to stuff I'm doing ;), on
reboot often the disk would appear to be in a coherent state, but from
an earlier point of view, sometimes hours (?!) precrash. Disk
amnesia. Edits would be gone, object files would no longer exist,
etc. Sometimes a file would be trashed. I became more serious about
checking in things I wanted to keep to a network CVS repository.
Previously in 2.1.xxx I _never_ had to manually run fsck after a
crash. I didn't loose data often either. Maybe I was lucky, maybe
the linux gods were smiling on me, but I did a lot of reboots, and the
only cost was the automated fsck on boot.
Now I get to run fsck on the disk far more frequently, maybe once a
month, post crash. Now I know that's still far cry from NT, where I
can loose the _disk_ _each_ crash, but who wants that as a standard?
I assume that there has been some change (either within the kernel or
in the RH 6.0 distro I have) which no longer causes dirty pages to be
written out unless there's a specific reason to, eg running sync. I
know that there's been a bunch of changes in the disk io area, around
2.3.9 as I recall, merging buffer and page table / cache functions.
Or perhaps I've got something mis-configured.
Is there a way (other than croning sync) to limit my exposure on disk
writes?
What was the rational for the change if one was intended?
Thanks.
-Richard
--
Richard Dynes
rdynes@varcom.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/