[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: The FC10 ext2 experiment has ended -- don't try this



Bill Davidsen wrote:
Robert Moskowitz wrote:
This is on an ASUS Eee PC netbook with 512Mb memory and a 4Gb SSD and a 8Gb SD card for more disk capacity (4Gb is simply not enough to even install FC10).

Only /home was on an ext3 partition. /boot, /, and /var were all ext2 partitions.

And things had been running well for a month or so. Regularly on cold boot, it would do an auto check on one of the ext2 partitions and reboot. Looked good....

Last night I suspended the box while plugged into AC; this worked well and it came right out of suspension this morning (on battery I use hibernate to save the battery). I then did a yum update.

The yum update was taking some time and the screensaver kicked in. For some reason this triggered the box to go into suspension (it never went into hibernation when I hibernated then yum update). I 'knew' bad things were going to happen to suspend right in the middle of a yum update...

You might look at your screensaver settings WRT power saving, you may have shot yourself in that foot.

I **thought** I had it set right. I suspect that because my last operation was a suspend, when it came out of suspend, something was left so when the timer kicked off (I had it set for 20min) for keyb/mouse inactivity, it suspended again instead of just going into screensaver.

Of course system the system was hosed, there was no way to find out, and I don't think I want to try again!

Using ext3 is hard on battery life.

I know. Even with noatime, you still have more disk activities. Of course, since my 'drives' are SSD and SD, this might not be such a difference (ext2 power usage vs ext3).
Also note there have been recent discussions of ext4 behavior if you shut down hard after writing and before the timer has physically written your data. I don't have details here, so I don't want to spread FUD, but a kernel patch was discussed, and the behavior as described does sound somewhat dangerous for laptop operation.

Okay, the URL was in my history, make you own evaluation:
http://www.h-online.com/open/Possible-data-loss-in-Ext4--/news/112821

I'm not going to depend on prayer to save my data until 2.6.30 is out and the fix is tested, not do I agree with Ted T'so that the applications should be fixed.

Is Ted on this list? I did not catch his ID. We worked together in the IETF 10 years ago.

IMHO any sequence of legal system calls should not zero out the files written..., certainly rewriting a block in a file is a pretty normal database operation, and should work at least as well as ext3.
Sure enough, inodes broken all over the place. Could NOT recover. Fortunately, there was nothing lost other than time.

I am rebuilding the system right now. I am keeping the partitions as I had them, but they will all be ext3 and I will change fstab for noatime...



[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]