Mario Charest
08/24/2010 2:40 PM
post64206
|
There is a partition of 100Meg, with 60 megs of files in it. An upgrade procedure tries to overwrite all of the 60 megs
of files with a newwer version, all with a single cp -r command. However it fails on a disk full error. Is that a side
effect of the way the fs works.
By the way this is on SSD and we have to use sync=none otherwise with sync=optional it doesn`t work at all ( IO error ).
I`m not using the qnx6fs for it`s new power fail safe feature but rather for the ability to support more then 2 boot
images.
|
|
|
Mate Szarvas
08/25/2010 2:07 AM
post64260
|
Mario, are you doing any control of snapshots, such as disabling
snapshots or changing the automatic snapshots interval (10 seconds by
default)?
My understanding is that if you try to write 60 megs in a single
transaction (without any snapshots being taken in the middle) then you
would need 120 megs of disk space because the filesystem maintains two
copies until the next snapshot.
If your SSD can take 40 megs faster than 10 secs then you may need to
split the cp -r and issue a number of sync-s in the middle to reclaim
area or, alternatively, reduce the snapshot period of the filesystem.
If "cp -r" filling in the empty space takes longer than 10 seconds (and
you did not tinker with snapshot periods) than I'd expect an automatic
snapshot to happen and free up space. In this case I cannot exaplain
what you see and you'll need to get input from someone with more
expertise.
--
Mate
-----Original Message-----
From: Mario Charest [mailto:community-noreply@qnx.com]
Sent: August 25, 2010 3:40 AM
To: general-filesystems
Subject: qnx6fs space usage
There is a partition of 100Meg, with 60 megs of files in it. An upgrade
procedure tries to overwrite all of the 60 megs of files with a newwer
version, all with a single cp -r command. However it fails on a disk
full error. Is that a side effect of the way the fs works.
By the way this is on SSD and we have to use sync=none otherwise with
sync=optional it doesn`t work at all ( IO error ).
I`m not using the qnx6fs for it`s new power fail safe feature but rather
for the ability to support more then 2 boot images.
_______________________________________________
General
http://community.qnx.com/sf/go/post64206
|
|
|
Keith Garvin
08/25/2010 2:50 AM
post64261
|
Hello Mario, Mate,
Mate, your understanding is correct. If the sync=none is set, then
there will be no timed snapshots and the operation you describe below
will likely fail with a disk full message just as you describe. As Mate
suggested changing your setup to perform frequent sync() of the file
system will force a snapshot.
I would be concerned about the IO errors you are seeing when
sync=optional is used. If you can provide a sloginfo output after the
failure case we might be able to see what is going on. Load the block
device driver with blk verbose=cvbf option so we can get more detailed
information in the slog.
Keith
-----Original Message-----
From: Mate Szarvas [mailto:community-noreply@qnx.com]
Sent: August 24, 2010 11:08 PM
To: general-filesystems
Subject: RE: qnx6fs space usage
Mario, are you doing any control of snapshots, such as disabling
snapshots or changing the automatic snapshots interval (10 seconds by
default)?
My understanding is that if you try to write 60 megs in a single
transaction (without any snapshots being taken in the middle) then you
would need 120 megs of disk space because the filesystem maintains two
copies until the next snapshot.
If your SSD can take 40 megs faster than 10 secs then you may need to
split the cp -r and issue a number of sync-s in the middle to reclaim
area or, alternatively, reduce the snapshot period of the filesystem.
If "cp -r" filling in the empty space takes longer than 10 seconds (and
you did not tinker with snapshot periods) than I'd expect an automatic
snapshot to happen and free up space. In this case I cannot exaplain
what you see and you'll need to get input from someone with more
expertise.
--
Mate
-----Original Message-----
From: Mario Charest [mailto:community-noreply@qnx.com]
Sent: August 25, 2010 3:40 AM
To: general-filesystems
Subject: qnx6fs space usage
There is a partition of 100Meg, with 60 megs of files in it. An upgrade
procedure tries to overwrite all of the 60 megs of files with a newwer
version, all with a single cp -r command. However it fails on a disk
full error. Is that a side effect of the way the fs works.
By the way this is on SSD and we have to use sync=none otherwise with
sync=optional it doesn`t work at all ( IO error ).
I`m not using the qnx6fs for it`s new power fail safe feature but rather
for the ability to support more then 2 boot images.
_______________________________________________
General
http://community.qnx.com/sf/go/post64206
_______________________________________________
General
http://community.qnx.com/sf/go/post64260
|
|
|
Mario Charest
|
Re: RE: qnx6fs space usage
|
Mario Charest
08/26/2010 11:10 AM
post64588
|
Re: RE: qnx6fs space usage
> Hello Mario, Mate,
>
> Mate, your understanding is correct. If the sync=none is set, then
> there will be no timed snapshots and the operation you describe below
> will likely fail with a disk full message just as you describe. As Mate
> suggested changing your setup to perform frequent sync() of the file
> system will force a snapshot.
>
> I would be concerned about the IO errors you are seeing when
> sync=optional is used. If you can provide a sloginfo output after the
> failure case we might be able to see what is going on. Load the block
> device driver with blk verbose=cvbf option so we can get more detailed
> information in the slog.
>
There it is, check at the bottom. devb-eide was restarted a few time as seen in the log.
The device is SSD OCZ Vextex. On earlier model of SSD ( ScanDisk ) the same image works fine.
The same image installed
Time Sev Major Minor Args
Aug 26 14:34:22 6 17 0 intrinfo size = 320 entry size = 64 count = 5
Aug 26 14:34:22 6 17 0 base = 0x00000000 num = 7 cascade = 07fffffff intr 48
Aug 26 14:34:22 6 17 0 base = 0x00000007 num = 1 cascade = 07fffffff intr 55
Aug 26 14:34:22 6 17 0 base = 0x00000008 num = 7 cascade = 07fffffff intr 56
Aug 26 14:34:22 6 17 0 base = 0x0000000f num = 1 cascade = 07fffffff intr 63
Aug 26 14:34:22 6 17 0 base = 0x80000000 num = 2 cascade = 07fffffff intr 2
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e - Vend 8086 - Class 1018a - Addr 0 - Size 8
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e - Vend 8086 - Class 1018a - Addr 0 - Size 4
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e - Vend 8086 - Class 1018a - Addr 0 - Size 8
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e - Vend 8086 - Class 1018a - Addr 0 - Size 4
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e - Vend 8086 - Class 1018a - Addr ffa0 - Size 10
Aug 26 14:34:22 5 14 0 tcpip starting
Aug 26 14:34:22 3 14 0 Using pseudo random generator. See "random" option
Aug 26 14:34:22 7 15 0 qnet(L4): qnet_birth(): qnet_init() - calling
Aug 26 14:34:22 2 19 0 devb-eide 1.00A (Jul 9 2010 13:54:17)
Aug 26 14:34:22 2 5 0 libcam.so (Jul 9 2010 13:53:08) bver 6040207
Aug 26 14:34:22 2 19 0 eide_identify_devices: Intel 63XXESB IDE vid 0x8086, did 0x269e, class 0x1018a rev 9,
busno 0, dfunc f9
Aug 26 14:34:22 2 19 0 eide_identify_devices: cmd_addr 0x1f0, cntl_addr 0x3f4, irq 0xe, chnl 0, udma 5, mdma 2
, sdma 0, pio 4
Aug 26 14:34:22 2 19 0 eide_identify_devices: probing dev 0 status 7c, error 7c
Aug 26 14:34:22 2 19 0 eide_identify_devices: probing dev 1 status 0, error 0
Aug 26...
View Full Message
|
|
|
Keith Garvin
|
RE: RE: qnx6fs space usage
|
Keith Garvin
08/26/2010 11:24 PM
post64714
|
RE: RE: qnx6fs space usage
Well this means the problem is not in the file system :) It looks like
the drive is actually returning not ready error condition after an
attempt to send the sync command, but a brief inspection of the code
does not reveal more details.
Get the verbosity of the devb-eide driver up to get a better clue as to
why it's failing. There are several options to the driver that may help
to better identify the root of the problem.
Keith
-----Original Message-----
From: Mario Charest [mailto:community-noreply@qnx.com]
Sent: August 26, 2010 8:10 AM
To: general-filesystems
Subject: Re: RE: qnx6fs space usage
> Hello Mario, Mate,
>
> Mate, your understanding is correct. If the sync=none is set, then
> there will be no timed snapshots and the operation you describe below
> will likely fail with a disk full message just as you describe. As
Mate
> suggested changing your setup to perform frequent sync() of the file
> system will force a snapshot.
>
> I would be concerned about the IO errors you are seeing when
> sync=optional is used. If you can provide a sloginfo output after the
> failure case we might be able to see what is going on. Load the block
> device driver with blk verbose=cvbf option so we can get more detailed
> information in the slog.
>
There it is, check at the bottom. devb-eide was restarted a few time as
seen in the log.
The device is SSD OCZ Vextex. On earlier model of SSD ( ScanDisk ) the
same image works fine.
The same image installed
Time Sev Major Minor Args
Aug 26 14:34:22 6 17 0 intrinfo size = 320 entry size = 64
count = 5
Aug 26 14:34:22 6 17 0 base = 0x00000000 num = 7 cascade =
07fffffff intr 48
Aug 26 14:34:22 6 17 0 base = 0x00000007 num = 1 cascade =
07fffffff intr 55
Aug 26 14:34:22 6 17 0 base = 0x00000008 num = 7 cascade =
07fffffff intr 56
Aug 26 14:34:22 6 17 0 base = 0x0000000f num = 1 cascade =
07fffffff intr 63
Aug 26 14:34:22 6 17 0 base = 0x80000000 num = 2 cascade =
07fffffff intr 2
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e -
Vend 8086 - Class 1018a - Addr 0 - Size 8
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e -
Vend 8086 - Class 1018a - Addr 0 - Size 4
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e -
Vend 8086 - Class 1018a - Addr 0 - Size 8
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e -
Vend 8086 - Class 1018a - Addr 0 - Size 4
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e -
Vend 8086 - Class 1018a - Addr ffa0 - Size 10
Aug 26 14:34:22 5 14 0 tcpip starting
Aug 26 14:34:22 3 14 0 Using pseudo random generator. See
"random" option
Aug 26 14:34:22 7 15 0 qnet(L4): qnet_birth():...
View Full Message
|
|
|
Mario Charest
|
RE: RE: qnx6fs space usage
|
Mario Charest
08/27/2010 3:46 PM
post64877
|
RE: RE: qnx6fs space usage
-----Message d'origine-----
De : Keith Garvin [mailto:community-noreply@qnx.com]
Envoyé : 26 août 2010 23:25
À : general-filesystems
Objet : RE: RE: qnx6fs space usage
> Well this means the problem is not in the file system :) It looks like the drive is
> actually returning not ready error condition after an attempt to send the sync
> command, but a brief inspection of the code does not reveal more details.
>Get the verbosity of the devb-eide driver up to get a better clue as
>to why it's failing. There are several options to the driver that may
>help to better identify the root of the problem.
I`ve turn every verbose option I could find. I've discovered there is a more recent firmware for these devices. I will
try it out Monday.
Time Sev Major Minor Args
Aug 27 19:29:33 6 17 0 intrinfo size = 320 entry size = 64 count = 5
Aug 27 19:29:33 6 17 0 base = 0x00000000 num = 7 cascade = 07fffffff intr 48
Aug 27 19:29:33 6 17 0 base = 0x00000007 num = 1 cascade = 07fffffff intr 55
Aug 27 19:29:33 6 17 0 base = 0x00000008 num = 7 cascade = 07fffffff intr 56
Aug 27 19:29:33 6 17 0 base = 0x0000000f num = 1 cascade = 07fffffff intr 63
Aug 27 19:29:33 6 17 0 base = 0x80000000 num = 2 cascade = 07fffffff intr 2
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 6 17 0 scan_device exiting
Aug 27 19:29:33 3 17 0 Range check failed (IO) - Dev 269e - Vend 8086 - Class 1018a - Addr 0 - Size 8
Aug 27 19:29:33 3 17 0 Range check failed (IO) - Dev 269e - Vend 8086 - Class 1018a - Addr 0 - Size 4
Aug 27 19:29:33 3 17 0 Range check failed (IO) - Dev 269e - Vend 8086 - Class 1018a - Addr 0 - Size 8
Aug 27 19:29:33 3 17 0 Range check failed (IO) - Dev 269e - Vend 8086 - Class 1018a - Addr 0 - Size 4
Aug 27 19:29:33 3 17 0 Range check failed (IO) - Dev 269e - Vend 8086 - Class 1018a - Addr ffa0 - Size 10
Aug 27 19:29:33 5 14 0 tcpip starting
Aug 27 19:29:33 3 14 0 Using pseudo random generator. See "random" option
Aug 27 19:29:33 7 15 0 qnet(L4): qnet_birth(): qnet_init() - calling
Aug 27 19:29:33 2 19 0 devb-eide 1.00A (Jul 9 2010 13:54:17)
Aug 27 19:29:33 2 5 0 libcam.so (Jul 9 2010 13:53:08) bver 6040207
Aug 27 19:29:33 2 19 0 eide_identify_devices: Intel 63XXESB IDE vid 0x8086, did 0x269e, class 0x1018a rev 9,
busno 0, dfunc f9
Aug 27 19:29:33 2 19 0 eide_identify_devices: cmd_addr 0x1f0, cntl_addr 0x3f4, irq 0xe, chnl 0, udma 5, mdma 2
, sdma 0, pio 4
Aug 27 19:29:33 2 19 0 eide_identify_devices: probing dev 0 status 0, error 0
Aug 27 19:29:33 2 19 0 eide_identify_devices: probing dev 1 status 0, error 0
Aug 27 19:29:33 2 19 0 eide_identify_devices: Intel 63XXESB IDE vid 0x8086, did 0x269e, class 0x1018a rev 9,
busno 0, dfunc f9
Aug 27 19:29:33 2 ...
View Full Message
|
|
|
Mario Charest
|
RE: RE: qnx6fs space usage
|
Mario Charest
08/31/2010 2:26 PM
post65284
|
RE: RE: qnx6fs space usage
Upgrading the flashdisk`s fimrware made the problem go away. For anyone reading this, model is OCZ Vertex and firmware
must be version 1.5
-----Message d'origine-----
De : Keith Garvin [mailto:community-noreply@qnx.com]
Envoyé : 26 août 2010 23:25
À : general-filesystems
Objet : RE: RE: qnx6fs space usage
Well this means the problem is not in the file system :) It looks like the drive is actually returning not ready error
condition after an attempt to send the sync command, but a brief inspection of the code does not reveal more details.
Get the verbosity of the devb-eide driver up to get a better clue as to why it's failing. There are several options to
the driver that may help to better identify the root of the problem.
Keith
-----Original Message-----
From: Mario Charest [mailto:community-noreply@qnx.com]
Sent: August 26, 2010 8:10 AM
To: general-filesystems
Subject: Re: RE: qnx6fs space usage
> Hello Mario, Mate,
>
> Mate, your understanding is correct. If the sync=none is set, then
> there will be no timed snapshots and the operation you describe below
> will likely fail with a disk full message just as you describe. As
Mate
> suggested changing your setup to perform frequent sync() of the file
> system will force a snapshot.
>
> I would be concerned about the IO errors you are seeing when
> sync=optional is used. If you can provide a sloginfo output after the
> failure case we might be able to see what is going on. Load the block
> device driver with blk verbose=cvbf option so we can get more detailed
> information in the slog.
>
There it is, check at the bottom. devb-eide was restarted a few time as seen in the log.
The device is SSD OCZ Vextex. On earlier model of SSD ( ScanDisk ) the same image works fine.
The same image installed
Time Sev Major Minor Args
Aug 26 14:34:22 6 17 0 intrinfo size = 320 entry size = 64
count = 5
Aug 26 14:34:22 6 17 0 base = 0x00000000 num = 7 cascade =
07fffffff intr 48
Aug 26 14:34:22 6 17 0 base = 0x00000007 num = 1 cascade =
07fffffff intr 55
Aug 26 14:34:22 6 17 0 base = 0x00000008 num = 7 cascade =
07fffffff intr 56
Aug 26 14:34:22 6 17 0 base = 0x0000000f num = 1 cascade =
07fffffff intr 63
Aug 26 14:34:22 6 17 0 base = 0x80000000 num = 2 cascade =
07fffffff intr 2
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 6 17 0 scan_device exiting
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e -
Vend 8086 - Class 1018a - Addr 0 - Size 8
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e -
Vend 8086 - Class 1018a - Addr 0 - Size 4
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e -
Vend 8086 - Class 1018a - Addr 0 - Size 8
Aug 26 14:34:22 3 17 0 Range check failed (IO) - Dev 269e -
Vend 8086 - Class 1018a - Addr 0 - Size 4
Aug...
View Full Message
|
|
|
|