Re: Transfering large files (was: Unison hangs on copy)

2011-07-30 Thread Victor Munoz
On Tue, Jul 26, 2011 at 03:24:11PM +0200, Claudius Hubig wrote:
 Victor Munoz vmu...@macul.ciencias.uchile.cl wrote:
 Both replicas are very large, and the few listed changes showed no
 large files, or so I thought. But it turns out I was wrong, and a 20M
 file was involved. I deleted the cache file, and reconstructed the
 mirror, and I finally discovered that unison hangs when this
 particular large files is involved. There are other large (and larger,
 4 times larger, for instance) files in the replicas, but they are
 identical, and unison doesn't complain. Only with this file, which
 differs. 
 
 So I tried rsyncinc another file, 23 M in size (with --progress
 option), and surprise, it stops when 11% transfer is reached. Tried
 with scp, and same magic number: 11% and it stops.
 
 What happens if you try to copy this file locally or on some other
 device or force reading it with, for example, dd (dd
 if=/path/to/yourfile of=/dev/null). Maybe the storage device on which
 it is located has some bad blocks. You could also try to run scp
 within strace to see where or what hangs.
 

I tried various combinations of this. I tried several different files,
all above 1M, none was fully transfered by rsync or scp :-( Looks like
an issue with keeping the connection alive. Using dd doesn't give
problems. scp within strace shows, at the end, messages like this:

write(6, )\377K\373)\377X\373#\377\373
\377n\373$\377\206\373\\377\231\373\36\377\257\373\\377\323\373...,
16384) = -1 EAGAIN (Resource temporarily unavailable)
poll([{fd=6, events=POLLOUT}], 1, -1)   = ? ERESTART_RESTARTBLOCK (To
be restarted)
--- SIGALRM (Alarm clock) @ 0 (0) ---

By this time the copy is stalled. 

Victor




-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: 
http://lists.debian.org/20110730230655.gb18...@llacolen.ciencias.uchile.cl



Re: Transfering large files (was: Unison hangs on copy)

2011-07-27 Thread Randy Kramer
On Tuesday 26 July 2011 09:28:49 am Victor Munoz wrote:
 Fine, thanks for the speculation and the interest.

You're welcome, and good luck!

Randy Kramer


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/201107271853.41032.rhkra...@gmail.com



Re: Transfering large files (was: Unison hangs on copy)

2011-07-26 Thread Randy Kramer
I've reordered some of your comments for my convenience in responding.

On Tuesday 26 July 2011 12:32:18 am Victor Munoz wrote:
 Thought you guys who helped me with my original question would like
 to know how this evolved.

Thank you!

 So I tried rsyncinc another file, 23 M in size (with --progress
 option), and surprise, it stops when 11% transfer is reached. Tried
 with scp, and same magic number: 11% and it stops.

I'm somewhat surprised, because when I was thinking of large files, I 
was thinking of CD image size files.  (I used to transfer these over a 
33 kbps modem, requiring about 65 hours per image, spread out over 
(typically) 5 nights.  Oh, the good old days ;-)

I don't really have any good thoughts to offer.  I'd be looking to make 
sure that no temporary storage areas have been filled (presumably /tmp 
on either machine), in fact, I'd probably do a df for all partitions on 
both machines and make sure all of them have plenty of space (a nice 
big multiple of 23 MB--I don't really know how much storage rsync, or 
scp need, but I'm just grasping at straws looking for possibilities.

While rsync may do some thinking (i.e., pause) during a 23 MB file 
transfer, I don't think scp would, but there still might be something 
that is timing out. 

All from me is just random speculation, however.  Good luck--I hope 
someone else will have some better suggestions for you.

Randy Kramer

 So now I think I have a problem with transfering large files. Somehow
 the upgrade caused this, because almost all previous transfers
 involved that 20M file, but now I am unable to copy this and other
 large files between both machines.

 I'm also losing my connection to the remote machine as I type this
 mail, so I have some kind of issue with the connection, and I really
 hope it is the the reason, and not the size of the files itself.





-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/201107260725.41630.rhkra...@gmail.com



Re: Transfering large files (was: Unison hangs on copy)

2011-07-26 Thread Claudius Hubig
Victor Munoz vmu...@macul.ciencias.uchile.cl wrote:
Both replicas are very large, and the few listed changes showed no
large files, or so I thought. But it turns out I was wrong, and a 20M
file was involved. I deleted the cache file, and reconstructed the
mirror, and I finally discovered that unison hangs when this
particular large files is involved. There are other large (and larger,
4 times larger, for instance) files in the replicas, but they are
identical, and unison doesn't complain. Only with this file, which
differs. 

So I tried rsyncinc another file, 23 M in size (with --progress
option), and surprise, it stops when 11% transfer is reached. Tried
with scp, and same magic number: 11% and it stops.

What happens if you try to copy this file locally or on some other
device or force reading it with, for example, dd (dd
if=/path/to/yourfile of=/dev/null). Maybe the storage device on which
it is located has some bad blocks. You could also try to run scp
within strace to see where or what hangs.

So now I think I have a problem with transfering large files. Somehow
the upgrade caused this, because almost all previous transfers
involved that 20M file, but now I am unable to copy this and other
large files between both machines. 

Is only this one file the problem or maybe other files as well? What
happens if you copy (not move!) that file to another location on your
drive and then try to scp it over? Is the content of the file thus
that you could send it via plain netcat? If so, did you try something
along the lines of:

destination# nc -l yourport | md5sum

source# dd if=/path/to/mysterious/file | nc otherhost.example.org yourport

Please note that I have not tested the commands given here, so you
might want to have a look at the manpages to make sure you got them correct.

Best regards,

Claudius
-- 
If only you had a personality instead of an attitude.
Please use GPG: ECB0C2C7 4A4C4046 446ADF86 C08112E5 D72CDBA4
http://chubig.net/ http://nightfall.org



-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: http://lists.debian.org/20110726152411.23a3d...@ares.home.chubig.net



Re: Transfering large files (was: Unison hangs on copy)

2011-07-26 Thread Victor Munoz
On Tue, Jul 26, 2011 at 07:25:41AM -0400, Randy Kramer wrote:
  So I tried rsyncinc another file, 23 M in size (with --progress
  option), and surprise, it stops when 11% transfer is reached. Tried
  with scp, and same magic number: 11% and it stops.
 
 I'm somewhat surprised, because when I was thinking of large files, I 
 was thinking of CD image size files.  (I used to transfer these over a 
 33 kbps modem, requiring about 65 hours per image, spread out over 
 (typically) 5 nights.  Oh, the good old days ;-)

Indeed, this is very frustrating, 20M is not a huge file, but it's the
size which is giving me trouble (ok, it could be less, haven't tried,
but I don't want to be too obsessive :-)

 
 I don't really have any good thoughts to offer.  I'd be looking to make 
 sure that no temporary storage areas have been filled (presumably /tmp 
 on either machine), in fact, I'd probably do a df for all partitions on 
 both machines and make sure all of them have plenty of space (a nice 
 big multiple of 23 MB--I don't really know how much storage rsync, or 
 scp need, but I'm just grasping at straws looking for possibilities.
 
 While rsync may do some thinking (i.e., pause) during a 23 MB file 
 transfer, I don't think scp would, but there still might be something 
 that is timing out. 

I thought exactly the same thing, that's why I tried scp after rsync.
One of the machines reports this:

df -h
FilesystemSize  Used Avail Use% Mounted on
/dev/sda1 178G  146G   24G  87% /
tmpfs1007M 0 1007M   0% /lib/init/rw
udev 1002M  172K 1002M   1% /dev
tmpfs1007M 0 1007M   0% /dev/shm

This is the remote machine I'm trying to synchronize with every night.
I will check at home the other one when I can. 


 
 All from me is just random speculation, however.  Good luck--I hope 
 someone else will have some better suggestions for you.

Fine, thanks for the speculation and the interest.

Victor



-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: 
http://lists.debian.org/20110726132849.gc11...@llacolen.ciencias.uchile.cl



Re: Transfering large files (was: Unison hangs on copy)

2011-07-26 Thread Victor Munoz
On Tue, Jul 26, 2011 at 03:24:11PM +0200, Claudius Hubig wrote:
 Victor Munoz vmu...@macul.ciencias.uchile.cl wrote:
 
 So I tried rsyncinc another file, 23 M in size (with --progress
 option), and surprise, it stops when 11% transfer is reached. Tried
 with scp, and same magic number: 11% and it stops.
 
 What happens if you try to copy this file locally or on some other
 device or force reading it with, for example, dd (dd
 if=/path/to/yourfile of=/dev/null). Maybe the storage device on which
 it is located has some bad blocks. You could also try to run scp
 within strace to see where or what hangs.

Ok, I will try this and your other suggestions when I'm home later in
the day, since I don't have access to that machine now.

 
 So now I think I have a problem with transfering large files. Somehow
 the upgrade caused this, because almost all previous transfers
 involved that 20M file, but now I am unable to copy this and other
 large files between both machines. 
 
 Is only this one file the problem or maybe other files as well? What
 happens if you copy (not move!) that file to another location on your
 drive and then try to scp it over? 

After I noticed unison had problems with this file, I took another
file, about the same size, in other directory, and did the rsync and
scp experiments with those files. So I don't think it is a problem
with that particular file or disk sector. But I will check your other
suggestions as soon as I can.

Thanks,

Victor


-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: 
http://lists.debian.org/20110726133425.gd11...@llacolen.ciencias.uchile.cl



Transfering large files (was: Unison hangs on copy)

2011-07-25 Thread Victor Munoz
On Fri, Jul 22, 2011 at 11:07:58PM -0400, Victor Munoz wrote:
 Today, I lost the ability to synchronize between two machines, one had
 lenny until today morning, and the other was sid. I was using
 apt-pinning to keep both unison versions compatible.

I few days ago I posted a problem with unison, which appeared after a
particular update I did. Someone suggested it could be a problem with
large files, and I said no, no large files involved. Ok, I should I
was wrong.

Both replicas are very large, and the few listed changes showed no
large files, or so I thought. But it turns out I was wrong, and a 20M
file was involved. I deleted the cache file, and reconstructed the
mirror, and I finally discovered that unison hangs when this
particular large files is involved. There are other large (and larger,
4 times larger, for instance) files in the replicas, but they are
identical, and unison doesn't complain. Only with this file, which
differs. 

So I tried rsyncinc another file, 23 M in size (with --progress
option), and surprise, it stops when 11% transfer is reached. Tried
with scp, and same magic number: 11% and it stops.

So now I think I have a problem with transfering large files. Somehow
the upgrade caused this, because almost all previous transfers
involved that 20M file, but now I am unable to copy this and other
large files between both machines. 

I'm also losing my connection to the remote machine as I type this
mail, so I have some kind of issue with the connection, and I really
hope it is the the reason, and not the size of the files itself.
Thought you guys who helped me with my original question would like to
know how this evolved. 

Regards,

Victor



-- 
To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org 
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org
Archive: 
http://lists.debian.org/20110726043218.ga7...@llacolen.ciencias.uchile.cl