Eric Sandeen wrote:
On 2/14/12 10:46 AM, Tom Crane wrote:
Dear XFS Support,
I am attempting to use xfs_fsr to defrag a 60TB FS but am getting some of the following errors;
'XFS_IOC_SWAPEXT failed: ino=xxxxxx: Invalid argument'. Most files defrag w/o problem. In an hour long run only 45/(45+6211) failed this way. Here is a example chunk of syslog from a run with fsr -v which includes the FS level reports.
Feb 14 15:49:13 store3 fsr[10917]: extents before:10 after:1 DONE ino=797765
Feb 14 15:49:13 store3 fsr[10917]: ino=797738
Feb 14 15:49:13 store3 fsr[10917]: extents before:9 after:1 DONE ino=797738
Feb 14 15:49:13 store3 fsr[10917]: ino=797749
Feb 14 15:49:14 store3 fsr[10917]: extents before:8 after:1 DONE ino=797749
Feb 14 15:49:14 store3 fsr[10917]: ino=797754
Feb 14 15:49:15 store3 fsr[10917]: extents before:8 after:1 DONE ino=797754
Feb 14 15:49:15 store3 fsr[10917]: ino=797728
Feb 14 15:49:17 store3 kernel: Filesystem dm-0: fs/xfs/xfs_dfrag.c: inode 0xc2c20 format is incompatible for exchanging.
Feb 14 15:49:17 store3 fsr[10917]: XFS_IOC_SWAPEXT failed: ino=797728: Invalid argument
Feb 14 15:49:17 store3 fsr[10917]: ino=797753
Feb 14 15:49:18 store3 kernel: Filesystem dm-0: fs/xfs/xfs_dfrag.c: inode 0xc2c39 format is incompatible for exchanging.
Feb 14 15:49:18 store3 fsr[10917]: XFS_IOC_SWAPEXT failed: ino=797753: Invalid argument
Feb 14 15:49:18 store3 fsr[10917]: ino=797740
Feb 14 15:49:20 store3 fsr[10917]: extents before:6 after:1 DONE ino=797740
Feb 14 15:49:20 store3 fsr[10917]: ino=797721
Feb 14 15:49:21 store3 fsr[10917]: extents before:5 after:1 DONE ino=797721
Feb 14 15:49:21 store3 fsr[10917]: ino=797720
Feb 14 15:49:22 store3 fsr[10917]: extents before:4 after:1 DONE ino=797720
Feb 14 15:49:22 store3 fsr[10917]: ino=797723
Feb 14 15:49:23 store3 fsr[10917]: extents before:4 after:1 DONE ino=797723
I have had a browse in the archive and can rule out an SElinux attribute difference (using xfs_io -c lsattr) between the problem files and the others. It is not an busy file problem either. I've rechecked with fuser and xfs_fsr -v on some of the individual files and always get the same error. xfs_bmaping the problem files afterwards shows they remain un-defragmented. Here is the output of xfs_bmap -v on the file with inode=797728.
EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET TOTAL FLAGS
0: [0..61439]: 81759234304..81759295743 38 (154865408..154926847) 61440 00011
1: [61440..127407]: 81959724544..81959790511 38 (355355648..355421615) 65968 00111
2: [127408..127791]: 81959790528..81959790911 38 (355421632..355422015) 384 01111
3: [127792..127807]: 81959790512..81959790527 38 (355421616..355421631) 16 01111
4: [127808..157695]: 81959791104..81959820991 38 (355422208..355452095) 29888 00111
5: [157696..186367]: 81959013120..81959041791 38 (354644224..354672895) 28672 00011
6: [186368..225039]: 81980197120..81980235791 38 (375828224..375866895) 38672 00111
I am running the latest (v.3.1.7) xfsprogs. My OS is SLC5 Linux with kernel details,
2.6.18-274.17.1.el5 #1 SMP Wed Jan 11 11:10:32 CET 2012 x86_64 x86_64 x86_64 GNU/Linux.
xfs_info reports the following for the FS,
xfs_info /dev/mapper/vg0-lvol0
meta-data=/dev/mapper/vg0-lvol0 isize=256 agcount=59, agsize=268435424 blks
= sectsz=512 attr=2
data = bsize=4096 blocks=15624994816, imaxpct=5
= sunit=32 swidth=128 blks
naming =version 2 bsize=4096 ascii-ci=0
log =internal bsize=4096 blocks=521728, version=2
= sectsz=512 sunit=32 blks, lazy-count=0
realtime =none extsz=524288 blocks=0, rtextents=0
Is this a known problem with xfs in this kernel? Any other information/tests that I can supply?
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=e09f98606dcc156de1146c209d45a0d6d5f51c3f
and
http://git.kernel.org/?p=fs/xfs/xfsprogs-dev.git;a=commitdiff;h=bdb041f58dc436dcb10b698ed8715fb889589b90
contain a lot of comments about what it is you're running into. The latter (the userspace change) should have made these less frequent.
If you're familiar with tracepoints, can you enable these and watch?
This should do it:
# mount -t debugfs none /sys/kernel/debug
# echo 1 > /sys/kernel/debug/tracing/tracing_enabled
# echo 1 > /sys/kernel/debug/tracing/events/xfs/xfs_swap_extent_before/enable
# echo 1 > /sys/kernel/debug/tracing/events/xfs/xfs_swap_extent_after/enable
Thanks for that tip but I've not tried this before and can't get it to
work. After mounting debugfs I don't have /sys/kernel/debug/tracing. ie,
find /sys/kernel/debug -print
/sys/kernel/debug
/sys/kernel/debug/usbmon
/sys/kernel/debug/usbmon/2s
/sys/kernel/debug/usbmon/2t
/sys/kernel/debug/usbmon/1s
/sys/kernel/debug/usbmon/1t
The .config file that comes with the pre-compiled kernel has the
following set in the kernel hacking section.
CONFIG_TRACE_IRQFLAGS_SUPPORT=y
CONFIG_MAGIC_SYSRQ=y
CONFIG_DEBUG_KERNEL=y
CONFIG_LOG_BUF_SHIFT=19
CONFIG_DETECT_SOFTLOCKUP=y
CONFIG_DETECT_HUNG_TASK=y
CONFIG_BOOTPARAM_HUNG_TASK_PANIC_VALUE=0
CONFIG_SCHEDSTATS=y
CONFIG_DEBUG_INFO=y
CONFIG_DEBUG_FS=y
CONFIG_DEBUG_LIST=y
CONFIG_BOOT_DELAY=y
CONFIG_SAMPLES=y
CONFIG_SAMPLE_MARKERS=m
CONFIG_SAMPLE_TRACEPOINTS=m
CONFIG_DEBUG_RODATA=y
CONFIG_DEBUG_STACKOVERFLOW=y
CONFIG_X86_DECODER_SELFTEST=y
A full copy of .config is on http://www.pp.rhul.ac.uk/~tcrane/.config
Is anything missing ?
Thanks
Tom Crane
<run your failing fsr run>
# cat /sys/kernel/debug/tracing/trace
and that should give us more info.
Thanks,
-Eric
Many thanks
Tom Crane
_______________________________________________
xfs mailing list
xfs@xxxxxxxxxxx
http://oss.sgi.com/mailman/listinfo/xfs
_______________________________________________
xfs mailing list
xfs@xxxxxxxxxxx
http://oss.sgi.com/mailman/listinfo/xfs
