* XFS filesystem reports as full though it isn't
@ 2008-05-16 20:27 Emmanuel Florac
2008-05-18 12:45 ` Emmanuel Florac
` (2 more replies)
0 siblings, 3 replies; 13+ messages in thread
From: Emmanuel Florac @ 2008-05-16 20:27 UTC (permalink / raw)
To: xfs
I have a 64 bits (x86_64) machine running Linux 2.6.22.19 with a 24TB
XFS filesystem. There are some 15TB of data on it. All is well, no
error except that I can't create a single file (touch foo : no space
left on device). I don't understand what can be going wrong...
History : this filesystem was extended (xfs_growfs) from 16TB to 24.
I've tought about all XFS problems I can remember of; first, inode
exhaustion :
cluster1:/proc/sys/fs# df -i
Sys. de fich. Inodes IUtil. ILib. %IUti. Mont?sur
/dev/sda1 0 0 0 - /
tmpfs 2058624 3 2058621 1% /lib/init/rw
udev 2058624 1445 2057179 1% /dev
tmpfs 2058624 1 2058623 1% /dev/shm
/dev/mapper/vg0-lv0 25379856384 4287 25379852097 1% /mnt/raid
OK so there's still plenty of inodes available here, that should be OK.
Then maybe we have something wrong elsewhere :
cluster1:/proc/sys/fs# xfs_db -r -c 'freesp -s -a 0' /dev/vg0/lv0
from to extents blocks pct
1 1 5 5 71.43
2 3 1 2 28.57
total free extents 6
total free blocks 7
average free extent size 1.16667
7 blocks free? But there are 9 TB of free space!
Here is the output from xfs_info /dev/vg0/lv0
meta-data=/dev/vg0/lv0 isize=256 agcount=47,
agsize=137245616 blks = sectsz=512 attr=0
data = bsize=4096 blocks=6344964096,
imaxpct=25 = sunit=16 swidth=32 blks,
unwritten=1 naming =version 2 bsize=4096
log =internal bsize=4096 blocks=32768, version=1
= sectsz=512 sunit=0 blks
realtime =none extsz=131072 blocks=0, rtextents=0
I fail to see nothing special there however.
The only significant thing I see is that the FS is really close to 16
TB of allocated data (15.7TB). I tried mounting it with "inode64"
option with no more loving.
Any help would be tremendously welcome. Users are starving for more
space and they can't even create a single file!
--
--------------------------------------------------
Emmanuel Florac www.intellique.com
--------------------------------------------------
^ permalink raw reply [flat|nested] 13+ messages in thread* Re: XFS filesystem reports as full though it isn't 2008-05-16 20:27 XFS filesystem reports as full though it isn't Emmanuel Florac @ 2008-05-18 12:45 ` Emmanuel Florac 2008-05-18 13:18 ` Christoph Hellwig 2008-05-19 11:31 ` Christian Røsnes 2 siblings, 0 replies; 13+ messages in thread From: Emmanuel Florac @ 2008-05-18 12:45 UTC (permalink / raw) Cc: xfs Le Fri, 16 May 2008 22:27:55 +0200 vous écriviez: > History : this filesystem was extended (xfs_growfs) from 16TB to 24. > One important thing I forgot to mention : some time before I have extended the FS from 16 to 24 TB, the FS got filled up commpletely (186MB free!!!). I've asked the user to make some room to avoid problems, so they freed 1.6 TB in a couple of days before growing up the FS. Is it possible that something got wrong with the filesystem being too filled? -- -------------------------------------------------- Emmanuel Florac www.intellique.com -------------------------------------------------- ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: XFS filesystem reports as full though it isn't 2008-05-16 20:27 XFS filesystem reports as full though it isn't Emmanuel Florac 2008-05-18 12:45 ` Emmanuel Florac @ 2008-05-18 13:18 ` Christoph Hellwig 2008-05-18 13:21 ` Christoph Hellwig 2008-05-19 11:31 ` Christian Røsnes 2 siblings, 1 reply; 13+ messages in thread From: Christoph Hellwig @ 2008-05-18 13:18 UTC (permalink / raw) To: Emmanuel Florac; +Cc: xfs On Fri, May 16, 2008 at 10:27:55PM +0200, Emmanuel Florac wrote: > cluster1:/proc/sys/fs# xfs_db -r -c 'freesp -s -a 0' /dev/vg0/lv0 > from to extents blocks pct > 1 1 5 5 71.43 > 2 3 1 2 28.57 > total free extents 6 > total free blocks 7 > average free extent size 1.16667 > > 7 blocks free? But there are 9 TB of free space! The -a 0 means you're only looking at the first AG, which given the history of the filesystem has a fair chance to be full. To get the summary for all AG do a xfs_db -r -c 'freesp -s' /dev/vg0/lv0 No idea yet on what could cause your problem, sorry. > The only significant thing I see is that the FS is really close to 16 > TB of allocated data (15.7TB). I tried mounting it with "inode64" > option with no more loving. Did you remount it with that option as in mount -o remount or did you completely unmount it and mount it again? Unfortunately none of the xfs-specific mount options can be reset with remount yet, and it doesn't return an error either (both are added to my todo list now) ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: XFS filesystem reports as full though it isn't 2008-05-18 13:18 ` Christoph Hellwig @ 2008-05-18 13:21 ` Christoph Hellwig 2008-05-18 13:42 ` Emmanuel Florac 0 siblings, 1 reply; 13+ messages in thread From: Christoph Hellwig @ 2008-05-18 13:21 UTC (permalink / raw) To: Emmanuel Florac; +Cc: xfs On Sun, May 18, 2008 at 09:18:55AM -0400, Christoph Hellwig wrote: > > The only significant thing I see is that the FS is really close to 16 > > TB of allocated data (15.7TB). I tried mounting it with "inode64" > > option with no more loving. > > Did you remount it with that option as in mount -o remount or did you > completely unmount it and mount it again? Unfortunately none of the > xfs-specific mount options can be reset with remount yet, and it doesn't > return an error either (both are added to my todo list now) Oh, and please post the output of /proc/self/mounts to see if all expected mount options are actualy seen by the filesystem. ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: XFS filesystem reports as full though it isn't 2008-05-18 13:21 ` Christoph Hellwig @ 2008-05-18 13:42 ` Emmanuel Florac 2008-05-19 20:10 ` Emmanuel Florac 0 siblings, 1 reply; 13+ messages in thread From: Emmanuel Florac @ 2008-05-18 13:42 UTC (permalink / raw) To: Christoph Hellwig; +Cc: xfs Le Sun, 18 May 2008 09:21:27 -0400 vous écriviez: > > Did you remount it with that option as in mount -o remount or did > > you completely unmount it and mount it again? Unfortunately none > > of the xfs-specific mount options can be reset with remount yet, > > and it doesn't return an error either (both are added to my todo > > list now) > IIRC I tried both, however I'll double check monday... > Oh, and please post the output of /proc/self/mounts to see if all > expected mount options are actualy seen by the filesystem. OK, thanks. -- -------------------------------------------------- Emmanuel Florac www.intellique.com -------------------------------------------------- ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: XFS filesystem reports as full though it isn't 2008-05-18 13:42 ` Emmanuel Florac @ 2008-05-19 20:10 ` Emmanuel Florac 2008-05-20 6:10 ` David Chinner 2008-05-20 10:19 ` Christoph Hellwig 0 siblings, 2 replies; 13+ messages in thread From: Emmanuel Florac @ 2008-05-19 20:10 UTC (permalink / raw) To: Christoph Hellwig; +Cc: xfs Le Sun, 18 May 2008 15:42:08 +0200 vous écriviez: > IIRC I tried both, however I'll double check monday... Well obviously I didn't try both, because it worked as expected when using "inode64". However there'ssomething slightly weird in the data layout: /dev/dm-0: AG Inodes IUsed IFree Use% 0 4288 4277 11 99% 1 0 0 0 0% 2 0 0 0 0% 3 0 0 0 0% 4 0 0 0 0% 5 0 0 0 0% 6 0 0 0 0% 7 0 0 0 0% 8 0 0 0 0% 9 0 0 0 0% 10 0 0 0 0% 11 0 0 0 0% 12 0 0 0 0% 13 0 0 0 0% 14 0 0 0 0% 15 0 0 0 0% 16 0 0 0 0% 17 0 0 0 0% 18 0 0 0 0% 19 0 0 0 0% 20 0 0 0 0% 21 0 0 0 0% 22 0 0 0 0% 23 0 0 0 0% 24 0 0 0 0% 25 0 0 0 0% 26 0 0 0 0% 27 0 0 0 0% 28 0 0 0 0% 29 0 0 0 0% 30 0 0 0 0% 31 0 0 0 0% 32 0 0 0 0% 33 0 0 0 0% 34 0 0 0 0% 35 0 0 0 0% 36 0 0 0 0% 37 0 0 0 0% 38 0 0 0 0% 39 0 0 0 0% 40 0 0 0 0% 41 0 0 0 0% 42 0 0 0 0% 43 0 0 0 0% 44 0 0 0 0% 45 0 0 0 0% 46 0 0 0 0% ALL 4288 4277 11 99% See how all inodes are in the same vg? is it OK? -- -------------------------------------------------- Emmanuel Florac www.intellique.com -------------------------------------------------- ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: XFS filesystem reports as full though it isn't 2008-05-19 20:10 ` Emmanuel Florac @ 2008-05-20 6:10 ` David Chinner 2008-05-20 10:17 ` Emmanuel Florac 2008-05-20 10:19 ` Christoph Hellwig 1 sibling, 1 reply; 13+ messages in thread From: David Chinner @ 2008-05-20 6:10 UTC (permalink / raw) To: Emmanuel Florac; +Cc: Christoph Hellwig, xfs On Mon, May 19, 2008 at 10:10:57PM +0200, Emmanuel Florac wrote: > Le Sun, 18 May 2008 15:42:08 +0200 vous écriviez: > > > IIRC I tried both, however I'll double check monday... > > Well obviously I didn't try both, because it worked as expected when > using "inode64". However there'ssomething slightly weird in the data > layout: > > /dev/dm-0: > AG Inodes IUsed IFree Use% > 0 4288 4277 11 99% > 1 0 0 0 0% > 2 0 0 0 0% ..... If there's no free space in AGs > 0 when you enable inode64, then you will still get enospc... > See how all inodes are in the same vg? is it OK? ^^ AG This is typical of inode32 allocation... Cheers, Dave. -- Dave Chinner Principal Engineer SGI Australian Software Group ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: XFS filesystem reports as full though it isn't 2008-05-20 6:10 ` David Chinner @ 2008-05-20 10:17 ` Emmanuel Florac 0 siblings, 0 replies; 13+ messages in thread From: Emmanuel Florac @ 2008-05-20 10:17 UTC (permalink / raw) To: David Chinner; +Cc: Christoph Hellwig, xfs Le Tue, 20 May 2008 16:10:47 +1000 David Chinner <dgc@sgi.com> écrivait: > If there's no free space in AGs > 0 when you enable inode64, then > you will still get enospc... > > > See how all inodes are in the same vg? is it OK? > ^^ AG > > This is typical of inode32 allocation... OK. I've mounted it "inode64" and apparently it worked fine. I started copying 6 more TB of data, I'll check how it goes this afternoon. -- ---------------------------------------- Emmanuel Florac | Intellique ---------------------------------------- ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: XFS filesystem reports as full though it isn't 2008-05-19 20:10 ` Emmanuel Florac 2008-05-20 6:10 ` David Chinner @ 2008-05-20 10:19 ` Christoph Hellwig 2008-05-20 12:39 ` Emmanuel Florac 1 sibling, 1 reply; 13+ messages in thread From: Christoph Hellwig @ 2008-05-20 10:19 UTC (permalink / raw) To: Emmanuel Florac; +Cc: Christoph Hellwig, xfs On Mon, May 19, 2008 at 10:10:57PM +0200, Emmanuel Florac wrote: > Le Sun, 18 May 2008 15:42:08 +0200 vous ?criviez: > > > IIRC I tried both, however I'll double check monday... > > Well obviously I didn't try both, because it worked as expected when > using "inode64". I've submitted a patch that should give back an error when remounting with an option not supported in remount, that should fix this little oversight in the future. Thanks for the report! ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: XFS filesystem reports as full though it isn't 2008-05-20 10:19 ` Christoph Hellwig @ 2008-05-20 12:39 ` Emmanuel Florac 0 siblings, 0 replies; 13+ messages in thread From: Emmanuel Florac @ 2008-05-20 12:39 UTC (permalink / raw) To: Christoph Hellwig; +Cc: xfs Le Tue, 20 May 2008 06:19:18 -0400 Christoph Hellwig <hch@infradead.org> écrivait: > I've submitted a patch that should give back an error when remounting > with an option not supported in remount, that should fix this little > oversight in the future. Thanks for the report! Thanks for the patch :) -- ---------------------------------------- Emmanuel Florac | Intellique ---------------------------------------- ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: XFS filesystem reports as full though it isn't 2008-05-16 20:27 XFS filesystem reports as full though it isn't Emmanuel Florac 2008-05-18 12:45 ` Emmanuel Florac 2008-05-18 13:18 ` Christoph Hellwig @ 2008-05-19 11:31 ` Christian Røsnes 2008-05-19 11:48 ` Christian Røsnes 2 siblings, 1 reply; 13+ messages in thread From: Christian Røsnes @ 2008-05-19 11:31 UTC (permalink / raw) To: Emmanuel Florac; +Cc: xfs On Fri, May 16, 2008 at 10:27 PM, Emmanuel Florac <eflorac@intellique.com> wrote: > > I have a 64 bits (x86_64) machine running Linux 2.6.22.19 with a 24TB > XFS filesystem. There are some 15TB of data on it. All is well, no > error except that I can't create a single file (touch foo : no space > left on device). I don't understand what can be going wrong... > > History : this filesystem was extended (xfs_growfs) from 16TB to 24. > > > Here is the output from xfs_info /dev/vg0/lv0 > > meta-data=/dev/vg0/lv0 isize=256 agcount=47, > agsize=137245616 blks = sectsz=512 attr=0 > data = bsize=4096 blocks=6344964096, > imaxpct=25 = sunit=16 swidth=32 blks, > unwritten=1 naming =version 2 bsize=4096 > log =internal bsize=4096 blocks=32768, version=1 > = sectsz=512 sunit=0 blks > realtime =none extsz=131072 blocks=0, rtextents=0 > > I fail to see nothing special there however. > > The only significant thing I see is that the FS is really close to 16 > TB of allocated data (15.7TB). I tried mounting it with "inode64" > option with no more loving. > On my system I get "no space left on device" when I reach 99% full with about 20GB free space left on 2TB partitions. I also use sunit and swidth for the data section of the xfs filesystem, and it could be that the XFS system cannot allocate space according to these parameters ? You seem to be around the original 16TB limit, and maybe it tries to allocate from this original disk layout ? (I'm no XFS expert, so please take my "theories" with a grain of salt) I recently had two "identical" partitions: A and B, where A was the master and B was the rsync copy. These had been written to for several years, and during that period they both have had the role as master. I suppose the disk usage layout was different. All of a sudden partition B reported "no space left on device", even though partition A contained the same data without any "no space left on device". What I did to get the B to copy the missing data from partition A was to: Temporarily move some data away from partition B, then run xfs_fsr on partition B, then move the data back. It brought the fragmentation down on partition B, and I could copy the missing data from partition B. Christian ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: XFS filesystem reports as full though it isn't 2008-05-19 11:31 ` Christian Røsnes @ 2008-05-19 11:48 ` Christian Røsnes 2008-05-19 20:39 ` Emmanuel Florac 0 siblings, 1 reply; 13+ messages in thread From: Christian Røsnes @ 2008-05-19 11:48 UTC (permalink / raw) To: Emmanuel Florac; +Cc: xfs On Mon, May 19, 2008 at 1:31 PM, Christian Røsnes <christian.rosnes@gmail.com> wrote: > On Fri, May 16, 2008 at 10:27 PM, Emmanuel Florac > <eflorac@intellique.com> wrote: >> >> I have a 64 bits (x86_64) machine running Linux 2.6.22.19 with a 24TB >> XFS filesystem. There are some 15TB of data on it. All is well, no >> error except that I can't create a single file (touch foo : no space >> left on device). I don't understand what can be going wrong... >> >> History : this filesystem was extended (xfs_growfs) from 16TB to 24. >> >> >> Here is the output from xfs_info /dev/vg0/lv0 >> >> meta-data=/dev/vg0/lv0 isize=256 agcount=47, >> agsize=137245616 blks = sectsz=512 attr=0 >> data = bsize=4096 blocks=6344964096, >> imaxpct=25 = sunit=16 swidth=32 blks, >> unwritten=1 naming =version 2 bsize=4096 >> log =internal bsize=4096 blocks=32768, version=1 >> = sectsz=512 sunit=0 blks >> realtime =none extsz=131072 blocks=0, rtextents=0 >> >> I fail to see nothing special there however. >> >> The only significant thing I see is that the FS is really close to 16 >> TB of allocated data (15.7TB). I tried mounting it with "inode64" >> option with no more loving. >> > > On my system I get "no space left on device" when I reach 99% full > with about 20GB free space left on 2TB partitions. > I also use sunit and swidth for the data section of the xfs > filesystem, and it could be that the XFS system cannot allocate > space according to these parameters ? You seem to be around the > original 16TB limit, and maybe it tries to allocate > from this original disk layout ? (I'm no XFS expert, so please take my > "theories" with a grain of salt) > > I recently had two "identical" partitions: A and B, where A was the > master and B was the rsync copy. Both these 2TB partitions (A and B) were 99% full. > These had been > written to for several years, and during that period they both have > had the role as master. I suppose the disk usage layout > was different. All of a sudden partition B reported "no space left on > device", even though partition A contained the > same data without any "no space left on device". What I did to get the > B to copy the missing data from partition A was to: > Temporarily move some data away from partition B, then run xfs_fsr on > partition B, then move the data back. > It brought the fragmentation down on partition B, and I could copy the > missing data from partition B. > Correction to last sentence: ... and I could copy the missing data _to_ partition B. Christian ^ permalink raw reply [flat|nested] 13+ messages in thread
* Re: XFS filesystem reports as full though it isn't 2008-05-19 11:48 ` Christian Røsnes @ 2008-05-19 20:39 ` Emmanuel Florac 0 siblings, 0 replies; 13+ messages in thread From: Emmanuel Florac @ 2008-05-19 20:39 UTC (permalink / raw) To: Christian Røsnes; +Cc: xfs Le Mon, 19 May 2008 13:48:01 +0200 vous écriviez: > Correction to last sentence: ... and I could copy the missing data > _to_ partition B. Another reminder that we should always keep our disks less than 95ù full :) -- -------------------------------------------------- Emmanuel Florac www.intellique.com -------------------------------------------------- ^ permalink raw reply [flat|nested] 13+ messages in thread
end of thread, other threads:[~2008-05-20 12:38 UTC | newest] Thread overview: 13+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2008-05-16 20:27 XFS filesystem reports as full though it isn't Emmanuel Florac 2008-05-18 12:45 ` Emmanuel Florac 2008-05-18 13:18 ` Christoph Hellwig 2008-05-18 13:21 ` Christoph Hellwig 2008-05-18 13:42 ` Emmanuel Florac 2008-05-19 20:10 ` Emmanuel Florac 2008-05-20 6:10 ` David Chinner 2008-05-20 10:17 ` Emmanuel Florac 2008-05-20 10:19 ` Christoph Hellwig 2008-05-20 12:39 ` Emmanuel Florac 2008-05-19 11:31 ` Christian Røsnes 2008-05-19 11:48 ` Christian Røsnes 2008-05-19 20:39 ` Emmanuel Florac
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox