Re: [Samba] 3.0.21c and big wmv or mpg files

2006-04-06 Thread Thomas Bork

Jeremy Allison wrote:


Well you know what strict allocate does, right ? It causes the space
for the file to be allocated on disk once the set fil allocation is
set from the Windows client. If the client then writes this data it'll
write everything twice. Why are you setting this ? You should not need
to set this in production unless you have some very specific needs.


The official documentation says:

Setting this to yes can help Samba return out of quota messages on 
systems that are restricting the disk quota of users.


This is suggesting me, that

strict allocate = yes

is the recommended setting for systems with quotas. That seems not the 
case...



der tom
--
To unsubscribe from this list go to the following URL and read the
instructions:  https://lists.samba.org/mailman/listinfo/samba


Re: [Samba] 3.0.21c and big wmv or mpg files

2006-04-06 Thread Jeremy Allison
On Thu, Apr 06, 2006 at 09:10:44PM +0200, Thomas Bork wrote:
 Jeremy Allison wrote:
 
 Well you know what strict allocate does, right ? It causes the space
 for the file to be allocated on disk once the set fil allocation is
 set from the Windows client. If the client then writes this data it'll
 write everything twice. Why are you setting this ? You should not need
 to set this in production unless you have some very specific needs.
 
 The official documentation says:
 
 Setting this to yes can help Samba return out of quota messages on 
 systems that are restricting the disk quota of users.
 
 This is suggesting me, that
 
 strict allocate = yes
 
 is the recommended setting for systems with quotas. That seems not the 
 case...

No, it is the case in that it will allow some (broken) Windows applications to 
get 
disk full messages at open time, rather than at write or close time. It 
shouldn't
(IMHO) be generically turned on just because you're using quotas unless you've
got specific need for it.

Just my 2 cents.

Jeremy.
-- 
To unsubscribe from this list go to the following URL and read the
instructions:  https://lists.samba.org/mailman/listinfo/samba


Re: [Samba] 3.0.21c and big wmv or mpg files

2006-04-05 Thread Thomas Bork

Tom Peters wrote:

While copying the wmv test file a have a cpu utilization of 95! 
percent (top) of smbd at the time, samba creates the sparse (?) file 
in the size needed for the test file.
I don't believe this happens in my case. I am not certain. I know that 
when it fails, Windows clients see a 0 bytes file with the name of the 
destination in the target share. You have to erase or overwrite it when 
you try again.


The high cpu utilization in our test case is caused of using quotas and 
an correspondending entry in smb.conf:


strict allocate = yes

Received a partially log from a user with a lot of

[2006/04/05 16:51:58, 3] smbd/error.c:error_packet(146)
  error packet at smbd/notify.c(55) cmd=160 (SMBnttrans) NT code 0x010c

Maybe this is problematical on smaller systems. Samba is running here 
on an VMWARE on top of XP (the test client) with P4 3,2 GHz:


One user tested without

strict allocate = yes

and can copy the file now.
Another user tested with more memory in his machine (celeron 1GHz, at 
first with 128 MB, then with 512 MB) and with


strict allocate = yes

and can copy the file now - but needs 50 minutes to copy this file...


der tom
--
To unsubscribe from this list go to the following URL and read the
instructions:  https://lists.samba.org/mailman/listinfo/samba


Re: [Samba] 3.0.21c and big wmv or mpg files

2006-04-05 Thread Jeremy Allison
On Wed, Apr 05, 2006 at 11:16:09PM +0200, Thomas Bork wrote:
 Tom Peters wrote:
 
 While copying the wmv test file a have a cpu utilization of 95! 
 percent (top) of smbd at the time, samba creates the sparse (?) file 
 in the size needed for the test file.
 I don't believe this happens in my case. I am not certain. I know that 
 when it fails, Windows clients see a 0 bytes file with the name of the 
 destination in the target share. You have to erase or overwrite it when 
 you try again.
 
 The high cpu utilization in our test case is caused of using quotas and 
 an correspondending entry in smb.conf:
 
 strict allocate = yes
 
 Received a partially log from a user with a lot of
 
 [2006/04/05 16:51:58, 3] smbd/error.c:error_packet(146)
   error packet at smbd/notify.c(55) cmd=160 (SMBnttrans) NT code 0x010c
 
 Maybe this is problematical on smaller systems. Samba is running here 
 on an VMWARE on top of XP (the test client) with P4 3,2 GHz:
 
 One user tested without
 
 strict allocate = yes
 
 and can copy the file now.
 Another user tested with more memory in his machine (celeron 1GHz, at 
 first with 128 MB, then with 512 MB) and with
 
 strict allocate = yes
 
 and can copy the file now - but needs 50 minutes to copy this file...

Well you know what strict allocate does, right ? It causes the space
for the file to be allocated on disk once the set fil allocation is
set from the Windows client. If the client then writes this data it'll
write everything twice. Why are you setting this ? You should not need
to set this in production unless you have some very specific needs.

Jeremy.
-- 
To unsubscribe from this list go to the following URL and read the
instructions:  https://lists.samba.org/mailman/listinfo/samba


Re: [Samba] 3.0.21c and big wmv or mpg files

2006-04-02 Thread Tom Peters

At 03:15 PM 4/1/2006 +0200, Thomas Bork wrote:

Tom Peters wrote:

Tell me, do you get this problem when you drag/n/drop a file into a 
folder on the samba share? And can you prevent this problem from 
occurring by the following procedure?


I cannot see the problem here but heared from 2 users of 3.0.21c with this 
problem.


Downloaded a file with 843 MB which reproducable triggered the error by 
one of the users - but not on my system.


We are using the _same_ samba version (built by me) on the same distribution:

http://www.eisfair.org/
http://www.pack-eis.de/index.php?p=samba

I'm the maintainer of this samba package and therefore interested in to 
correct this.


Click in the target window on the samba machine (this is on the XP 
desktop). Press F5 to refresh the view. Wait about a second. Immediately 
start your copy. When I do this, the errors, previously reported to the 
desktop and to my server log, don't occur.


The user copies the file with:

- right click to the local file and 'copy'
- right click in the share and 'insert'


This procedure is the same as drag-and-drop from WinXP point of view. In 
English I believe the options are 'copy' and 'paste' rather than 'copy' and 
'insert' but it's does the same as drag and drop in this context.


I've been told that it's a WinXP only issue, that it attempts to connect 
on ports 445 and 139 nearly simultaneously, and then proceeds to talk 
over whichever one answers first. I'm told that Win2k clients won't have 
this problem, and Win98 clients don't use port 445 so it doesn't arise there.


How did you adjusted your clients? On tab 'WINS'/Netbios properties (all 
translated from german):


TCP/IP properties/Default
or
TCP/IP properties/Enable Netbios over TCP/IP


Up to this moment, they had been TCP/IP properties/Default. I have 
changed it to Enable NetBIOS over TCP/IP.


While copying the wmv test file a have a cpu utilization of 95! percent 
(top) of smbd at the time, samba creates the sparse (?) file in the size 
needed for the test file.


I don't believe this happens in my case. I am not certain. I know that when 
it fails, Windows clients see a 0 bytes file with the name of the 
destination in the target share. You have to erase or overwrite it when you 
try again.


Maybe this is problematical on smaller systems. Samba is running here on 
an VMWARE on top of XP (the test client) with P4 3,2 GHz:


deveis # cat /proc/cpuinfo
processor   : 0
vendor_id   : GenuineIntel
cpu family  : 15
model   : 4
model name  : Intel(R) Pentium(R) 4 CPU 3.20GHz
stepping: 8
cpu MHz : 3193.393
cache size  : 0 KB
fdiv_bug: no
hlt_bug : no
f00f_bug: no
coma_bug: no
fpu : yes
fpu_exception   : yes
cpuid level : 5
wp  : yes
flags   : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca 
cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss pni ds_cpl

bogomips: 6370.09

By the user of the test file the sparse file with the correct size will be 
created but is complete empty...


I would need to force a failure of this type and examine the size of the 
target file on the Linux side. WinXP says its empty and zero-length.




der tom



[Philosophy] I loathe people who keep dogs. They are cowards who
haven't the guts to bite people themselves. --August Strindberg
--... ...--  -.. .  -. . --.- --.- -...
[EMAIL PROTECTED]   (remove nospam) N9QQB (amateur radio)
HEY YOU (loud shouting)  WEB ADDRESS http//www.mixweb.com/tpeters
43° 7' 17.2 N by 88° 6' 28.9 W,  Elevation 815',  Grid Square EN53wc
WAN/LAN/Telcom Analyst, Tech Writer, MCP, CCNA, Registered Linux User 385531



--
To unsubscribe from this list go to the following URL and read the
instructions:  https://lists.samba.org/mailman/listinfo/samba


Re: [Samba] 3.0.21c and big wmv or mpg files

2006-04-01 Thread Thomas Bork

Tom Peters wrote:

Tell me, do you get this problem when you drag/n/drop a file into a 
folder on the samba share? And can you prevent this problem from 
occurring by the following procedure?


I cannot see the problem here but heared from 2 users of 3.0.21c with 
this problem.


Downloaded a file with 843 MB which reproducable triggered the error by 
one of the users - but not on my system.


We are using the _same_ samba version (built by me) on the same 
distribution:


http://www.eisfair.org/
http://www.pack-eis.de/index.php?p=samba

I'm the maintainer of this samba package and therefore interested in to 
correct this.


Click in the target window on the samba machine (this is on the XP 
desktop). Press F5 to refresh the view. Wait about a second. Immediately 
start your copy. When I do this, the errors, previously reported to the 
desktop and to my server log, don't occur.


The user copies the file with:

- right click to the local file and 'copy'
- right click in the share and 'insert'

I've been told that it's a WinXP only issue, that it attempts to connect 
on ports 445 and 139 nearly simultaneously, and then proceeds to talk 
over whichever one answers first. I'm told that Win2k clients won't have 
this problem, and Win98 clients don't use port 445 so it doesn't arise 
there.


How did you adjusted your clients? On tab 'WINS'/Netbios properties (all 
translated from german):


TCP/IP properties/Default
or
TCP/IP properties/Enable Netbios over TCP/IP
?

While copying the wmv test file a have a cpu utilization of 95! percent 
(top) of smbd at the time, samba creates the sparse (?) file in the size 
needed for the test file.
Maybe this is problematical on smaller systems. Samba is running here on 
an VMWARE on top of XP (the test client) with P4 3,2 GHz:


deveis # cat /proc/cpuinfo
processor   : 0
vendor_id   : GenuineIntel
cpu family  : 15
model   : 4
model name  : Intel(R) Pentium(R) 4 CPU 3.20GHz
stepping: 8
cpu MHz : 3193.393
cache size  : 0 KB
fdiv_bug: no
hlt_bug : no
f00f_bug: no
coma_bug: no
fpu : yes
fpu_exception   : yes
cpuid level : 5
wp  : yes
flags   : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge 
mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss pni ds_cpl

bogomips: 6370.09

By the user of the test file the sparse file with the correct size will 
be created but is complete empty...



der tom
--
To unsubscribe from this list go to the following URL and read the
instructions:  https://lists.samba.org/mailman/listinfo/samba


Re: [Samba] 3.0.21c and big wmv or mpg files

2006-03-31 Thread Tom Peters

At 10:40 PM 3/29/2006 +0200, you wrote:

Hi @all,

are there any problems known with 3.0.21c and bigger video files (mpg and 
wmv greater than 700 MB)?


Received the information that these files cannot copied from XP to Samba 
(W2K is okay). Error message is the well known:


[2006/03/28 18:03:36, 0] lib/util_sock.c:get_peer_addr(1225)getpeername 
failed. Error was Transport endpoint is not connected


I chased this elusive problem for a year. I'm still running 3.09-2.3 but I 
see it on other versions.


Tell me, do you get this problem when you drag/n/drop a file into a folder 
on the samba share? And can you prevent this problem from occurring by the 
following procedure?


Click in the target window on the samba machine (this is on the XP 
desktop). Press F5 to refresh the view. Wait about a second. Immediately 
start your copy. When I do this, the errors, previously reported to the 
desktop and to my server log, don't occur.


I've been told that it's a WinXP only issue, that it attempts to connect on 
ports 445 and 139 nearly simultaneously, and then proceeds to talk over 
whichever one answers first. I'm told that Win2k clients won't have this 
problem, and Win98 clients don't use port 445 so it doesn't arise there.


Anyone have evidence to the contrary (so far)?

My thinking is that WinXP improperly responds to an attempt to open a 
conversation on a different port than the one Samba expects. Maybe the TCP 
stack on SuSE is more rigorous and respects the sequence numbers and 
considers a connection... ...a connection. Dunno.  Need to get busy with 
Ethereal maybe.


First thing I tried: In smb.com, add a line reading
smb ports = 139
which I hoped would tell it just don't reply on port 445 at all. This did 
not help.


Next thing I tried, rather a kludge,
iptables -I INPUT 1 -p tcp --dport 445 -j DROP

Hey guess what? I haven't had a single incidence of the error since.

My explanation, as near as I can figure out, of what the rule does:

iptables; firewall / packet filter
 -I INPUT 1 ; Insert into chain INPUT as rule #1
-p tcp --dport 445  ; a rule for packets whose protocol is tcp AND 
destination port is 445

 -j DROP; if matches rule, Jump to target DROP

The predefined target DROP is not another chain but actually means throw 
the packet away.
I'm allowing port 139 through (by default) instead of 445 because I still 
have the odd Win98 machine laying about.


That's my story, and I'm sticking to it.

-Tom






[Commentary] Despite of the cost of living, it remains popular.
--... ...--  -.. .  -. . --.- --.- -...
[EMAIL PROTECTED]   (remove nospam) N9QQB (amateur radio)
HEY YOU (loud shouting)  WEB ADDRESS http//www.mixweb.com/tpeters
43° 7' 17.2 N by 88° 6' 28.9 W,  Elevation 815',  Grid Square EN53wc
WAN/LAN/Telcom Analyst, Tech Writer, MCP, CCNA, Registered Linux User 385531



--
To unsubscribe from this list go to the following URL and read the
instructions:  https://lists.samba.org/mailman/listinfo/samba


Re: [Samba] 3.0.21c and big wmv or mpg files

2006-03-31 Thread Doug VanLeuven

Tom Peters wrote:

At 10:40 PM 3/29/2006 +0200, you wrote:

Hi @all,

are there any problems known with 3.0.21c and bigger video files (mpg 
and wmv greater than 700 MB)?


Received the information that these files cannot copied from XP to 
Samba (W2K is okay). Error message is the well known:


[2006/03/28 18:03:36, 0] 
lib/util_sock.c:get_peer_addr(1225)getpeername failed. Error was 
Transport endpoint is not connected


I chased this elusive problem for a year. I'm still running 3.09-2.3 but 
I see it on other versions.


Tell me, do you get this problem when you drag/n/drop a file into a 
folder on the samba share? And can you prevent this problem from 
occurring by the following procedure?


Click in the target window on the samba machine (this is on the XP 
desktop). Press F5 to refresh the view. Wait about a second. Immediately 
start your copy. When I do this, the errors, preterviously reported to the 
desktop and to my server log, don't occur.


I've been told that it's a WinXP only issue, that it attempts to connect 
on ports 445 and 139 nearly simultaneously, and then proceeds to talk 
over whichever one answers first. I'm told that Win2k clients won't have 
this problem, and Win98 clients don't use port 445 so it doesn't arise 
there.ou can take my word for it,


Anyone have evidence to the contrary (so far)?



Hi Tom,
If you can take my word for it,
I just drag  dropped a 2 Gig file from XP SP2 to
Samba version 3.0.22pre1-SVN-build-12802
on FC3 2.6.12-1.1381_FC3smp
without issue.  I do this pretty frequently moving vmware machines around
and organizing ghost images.

I run a 2003 AD domain, but the XP machine is just a workgroup
member of the domain. although samba is a domain member.  But I recall
doing this on an XP full domain member last year.

During the transfer:
PID Username  Group Machine
---
 9040   doug  doug  pine (192.168.200.14)

Service  pid machine   Connected at
---
public   9040   pine  Fri Mar 31 19:55:00 2006

Locked files:
PidDenyMode   Access  R/WOplock   SharePath   
Name
--
9040  DENY_ALL   0x30196 WRONLY EXCLUSIVE+BATCH  /home/public   doug/sda-s003.vmdk   Fri 
Mar 31 20:09:26 2006
9040  DENY_NONE  0x20089 RDONLY NONE /home/public   doug   Fri Mar 31 
19:55:05 2006
9040  DENY_NONE  0x11RDONLY NONE /home/public   doug   Fri Mar 31 
19:55:05 2006


And when done:
-rw-rw-rw-  1 doug doug 2125135872 Mar 15 11:56 /home/public/doug/sda-s003.vmdk

I did it once, deleted it, waited about 20 min and did it again.

Ports in use with the XP machine:
tcp0  0 192.168.200.25:445  192.168.200.14:1736 
ESTABLISHED
where samba is running on 192.168.200.25.

Perhaps relevant config option:
socket options = SO_KEEPALIVE IPTOS_LOWDELAY TCP_NODELAY SO_SNDBUF=65536 
SO_RCVBUF=65536

I did notice that when I first migrated from samba 2 to samba 3 back around 
3.0.9
at first the client machines continued to connect on port 139, but
gradually over a period of time that changed until the connections
are now nearly all port 445 and all the domain members list as IP numbers
instead of netbios names.

I've seen this behavior in windows clients where they remember connection
details and continue to use them until some event or loss of connectivity
causes the client to start over in the list.  Usually it learns the new
connection details at that time.

Regards, Doug
--
To unsubscribe from this list go to the following URL and read the
instructions:  https://lists.samba.org/mailman/listinfo/samba


[Samba] 3.0.21c and big wmv or mpg files

2006-03-29 Thread Thomas Bork

Hi @all,

are there any problems known with 3.0.21c and bigger video files (mpg 
and wmv greater than 700 MB)?


Received the information that these files cannot copied from XP to Samba 
(W2K is okay). Error message is the well known:


[2006/03/28 18:03:36, 0] lib/util_sock.c:get_peer_addr(1225)getpeername 
failed. Error was Transport endpoint is not connected


To eleminate network problems the client and the samba server were 
crossover connected...


Hope to get an testfile soon to reproduce and report this better.


der tom
--
To unsubscribe from this list go to the following URL and read the
instructions:  https://lists.samba.org/mailman/listinfo/samba


Re: [Samba] 3.0.21c and big wmv or mpg files

2006-03-29 Thread Jeremy Allison
On Wed, Mar 29, 2006 at 10:40:22PM +0200, Thomas Bork wrote:
 Hi @all,
 
 are there any problems known with 3.0.21c and bigger video files (mpg 
 and wmv greater than 700 MB)?
 
 Received the information that these files cannot copied from XP to Samba 
 (W2K is okay). Error message is the well known:
 
 [2006/03/28 18:03:36, 0] lib/util_sock.c:get_peer_addr(1225)getpeername 
 failed. Error was Transport endpoint is not connected
 
 To eleminate network problems the client and the samba server were 
 crossover connected...
 
 Hope to get an testfile soon to reproduce and report this better.

We need to know who dropped the TCP connection and why...
Test and logfile needed I'm afraid. What platform are
you running Samba on ?

Jeremy.
-- 
To unsubscribe from this list go to the following URL and read the
instructions:  https://lists.samba.org/mailman/listinfo/samba