You are not logged in.
I've encountered a problem when migrated two of my partitions from ext3 to ext4. /home partition was converted with
tune2fs -O extents,uninit_bg,dir_index /dev/the_partition
fsck -fDp /dev/the_partition
the other was just reformatted
mkfs.ext4 /dev/sdxY
here are mount options for both of those partitions
# mount
/dev/sda5 on /home type ext4 (rw,noatime,auto_da_alloc,journal_checksum)
/dev/sda8 on /var/storage type ext4 (rw,noatime,auto_da_alloc,journal_checksum,commit=20)
and fstab for them
# cat /etc/fstab
# <file system> <dir> <type> <options> <dump> <pass>
UUID=215039fd-04f4-4f18-b407-05941b12a464 /home ext4 defaults,noatime,auto_da_alloc,journal_checksum 0 1
UUID=c27bf3a2-6d6b-453a-94ea-e3fc378a79bc /var/storage ext4 defaults,noatime,nodiratime,auto_da_alloc,journal_checksum,commit=20 0 1
Each time I run
cp -vR /home/some_path /var/storage/some_path
I can see how the system hangs on some file. LED stops blinking, everything freezes, I can't even login to a new tty. Only if I had some tty open, I can run
sync
in that tty and the system comes alive again for some period of time, until next portion of data is read from the input and I have to run sync again.
I even found a workaround for those long cp -R operations:
while [ -d /proc/PID_OF_CP-R ] ; do sync ; sleep 10 ; done
Has anybody encountered the same nasty behavior?
PS. I have to mention I didn't have those problems when I was on ext3
Last edited by sealemar (2011-11-13 14:17:42)
Offline
Also this affects both console mode and X-mode.
What is more strange is that if I don't do any long cp -R, the system works ok. I can even play games, surf the Internet, do programming, etc. The system works well, until I run some long cp -R on any of ext4 partitions.
Offline
The problem was solved when I did a kernel compilation. That's not the only problem which was fixed. Check the thread about kworker constantly eating CPU time and about freezing during copying to/form a usb device. That one was fixed as well. During kernel configuration I noticed a lot of rubbish all-purpose-use options which were selected. A lot of options essential to me weren't selected. All in all custom kernel is not just fun to make, but really provides you with what you need and gives you a control over what needs to be configured and what others generic kernel users simply don't know.
Please, find my kernel config on pastebin if you're curious.
Here is my /etc/fstab
# <file system> <dir> <type> <options> <dump> <pass>
UUID=0da1c9be-1da9-4c24-a24f-e09c2f4b0fd5 /boot ext2 noauto,noatime 0 1
UUID=1bc57e15-f0ed-4d76-a831-300a109d5858 swap swap defaults 0 0
UUID=215039fd-04f4-4f18-b407-05941b12a464 /home ext4 defaults,noatime,nodiratime,auto_da_alloc,journal_checksum,commit=20 0 1
UUID=662265a7-ceac-49dd-9bca-0e89e69c0ba9 / reiserfs defaults,noatime 0 1
UUID=c27bf3a2-6d6b-453a-94ea-e3fc378a79bc /var/storage ext4 defaults,noatime,nodiratime,auto_da_alloc,journal_checksum,commit=20 0 1
Last edited by sealemar (2011-11-13 14:43:29)
Offline
@sealemar: noatime includes nodiratime.
zʇıɹɟʇıɹʞsuɐs AUR || Cycling in Budapest with a helmet camera || Revised log levels proposal: "FYI" "WTF" and "OMG" (John Barnette)
Offline
SanskritFritz, thank you. However, that's not obvious from the man page:
noatime
Do not update inode access times on this file system (e.g, for faster access on the news spool to speed up news servers).nodiratime
Do not update directory inode access times on this filesystem.
I've found some information upon the theme for those who are curious:
Does noatime imply nodiratime? - a discussion with a bit of code responsible for those options from touch_atime() in <fs/inode.c>
Last edited by sealemar (2011-11-15 03:15:41)
Offline