Re: [Samba] extremely high %CPU and TIME

2003-11-20 Thread Tom Ryan
Jeremy (and all),

just wanted to say that we moved (quicker than we wanted) to 3.0.x and 
everything is fine!

Thanks so much !

Tom

On Wed, 19 Nov 2003, Jeremy Allison wrote:

 On Wed, Nov 19, 2003 at 03:57:56PM -0500, Tom Ryan wrote:
  yes.. but I was having problems before with that enabled previously.
 
 The client behaviour changes once spoolss is enabled. If you're 
 claiming to see the same issue on the clients you need to restart
 the spooler service on them in order for them to notice the now
 open spoolss pipe.
 
 3.0.x has a *much* improved version of the printing code we use
 on the HP PSA (Print Server Appliance), so I can definately
 recommend it.
 
 Jeremy.
 

-- 
___
Tom RyanVoice: 856-225-6361
Consulting System Administrator   Fax: 856-969-7900
Rutgers School of Law - Camden

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


Re: [Samba] extremely high %CPU and TIME

2003-11-19 Thread Tom Ryan
I've kept looking for an answer to this to no avail.. any ideas?

strace -p  reveals

fcntl64(14, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=684, len=1}, 
0xb180) = 0
fcntl64(14, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=684, len=1}, 
0xb150) = 0
fcntl64(14, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=5676, 
len=1}, 0xb160) = 0
fcntl64(14, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=684, len=1}, 
0xb150) = 0
fcntl64(14, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=688, len=1}, 
0xb150) = 0
fcntl64(14, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=688, len=1}, 
0xb150) = 0
fcntl64(16, F_SETLKW64, {type=F_RDLCK, whence=SEEK_SET, start=528, len=1}, 
0xbfffda40) = 0
fcntl64(16, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=528, len=1}, 
0xbfffda40) = 0
time(NULL)  = 1069272774
fcntl64(16, F_SETLKW64, {type=F_RDLCK, whence=SEEK_SET, start=548, len=1}, 
0xbfffc8a0) = 0
fcntl64(16, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=548, len=1}, 
0xbfffc8e0) = 0

thanks!

On Thu, 9 Oct 2003, Tom Ryan wrote:

 Looking around on the net, I saw that someone else had a similar issue 
 when they were running all of their users behind terminal server. Our 
 users are all behind two NAT'd firewalls.
 
 Is it possible that there is some confusion going on? should I stop the 
 service and remove locking.tdb? (that's what 14 references right?)
 
 Thanks.
 
 tom
 
 On Wed, 8 Oct 2003, Tom Ryan wrote:
 
  2.4.20-19.8 on redhat 8
  
  strace reports 
  fcntl64(14, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=628, len=1}, 
  0xb150) = 0
  fcntl64(14, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=632, len=1}, 
  0xb150) = 0
  
  over and over again (with different start points)
  
  Tom
  
  On Wed, 8 Oct 2003, Jeremy Allison wrote:
  
   On Wed, Oct 08, 2003 at 02:29:31PM -0400, Tom Ryan wrote:
I have a samba box that has around 70 users on it at any given time.

When I run ps axu there are certain smbd processes (and usually always 
by the same login) that look like

username 26126 39.4  0.1  6652 2912 ?S13:44  17:12 smbd -D

any ideas why CPU is running at 39.4??
   
   What OS ? Try doing an strace on it to see what the process is doing.
   
   Jeremy.
   
  
  
 
 

-- 
___
Tom RyanVoice: 856-225-6361
Consulting System Administrator   Fax: 856-969-7900
Rutgers School of Law - Camden

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


Re: [Samba] extremely high %CPU and TIME

2003-11-19 Thread Jeremy Allison
On Wed, Nov 19, 2003 at 03:14:02PM -0500, Tom Ryan wrote:
 I've kept looking for an answer to this to no avail.. any ideas?
 
 strace -p  reveals
 
 fcntl64(14, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=684, len=1}, 
 0xb180) = 0
 fcntl64(14, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=684, len=1}, 
 0xb150) = 0

Ok, it's in a tdb call. Can you attach with gdb and get a stack backtrace ?

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


Re: [Samba] extremely high %CPU and TIME

2003-11-19 Thread Tom Ryan
Jeremy,

(gdb) bt
#0  0x08173176 in tdb_brlock ()
#1  0x0817349a in tdb_unlock ()
#2  0x0817517a in tdb_next_lock ()
#3  0x081752a6 in tdb_traverse ()
#4  0x08179de4 in print_queue_status ()
#5  0x0807a6de in api_DosPrintQGetInfo ()
#6  0x0807fdd8 in api_reply ()
#7  0x08077e03 in reply_trans ()
#8  0x080af1af in switch_message ()
#9  0x080af3a9 in construct_reply ()
#10 0x080af528 in process_smb ()
#11 0x080b01b4 in smbd_process ()
#12 0x0806ab4b in main ()
#13 0x420158f7 in __libc_start_main () from /lib/i686/libc.so.6
(gdb) 

On Wed, 19 Nov 2003, Jeremy Allison wrote:

 On Wed, Nov 19, 2003 at 03:14:02PM -0500, Tom Ryan wrote:
  I've kept looking for an answer to this to no avail.. any ideas?
  
  strace -p  reveals
  
  fcntl64(14, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=684, len=1}, 
  0xb180) = 0
  fcntl64(14, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=684, len=1}, 
  0xb150) = 0
 
 Ok, it's in a tdb call. Can you attach with gdb and get a stack backtrace ?
 
 Jeremy.
 

-- 
___
Tom RyanVoice: 856-225-6361
Consulting System Administrator   Fax: 856-969-7900
Rutgers School of Law - Camden

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


Re: [Samba] extremely high %CPU and TIME

2003-11-19 Thread Jeremy Allison
On Wed, Nov 19, 2003 at 03:23:31PM -0500, Tom Ryan wrote:
 Jeremy,
 
 (gdb) bt
 #0  0x08173176 in tdb_brlock ()
 #1  0x0817349a in tdb_unlock ()
 #2  0x0817517a in tdb_next_lock ()
 #3  0x081752a6 in tdb_traverse ()
 #4  0x08179de4 in print_queue_status ()
 #5  0x0807a6de in api_DosPrintQGetInfo ()
 #6  0x0807fdd8 in api_reply ()
 #7  0x08077e03 in reply_trans ()
 #8  0x080af1af in switch_message ()
 #9  0x080af3a9 in construct_reply ()
 #10 0x080af528 in process_smb ()
 #11 0x080b01b4 in smbd_process ()
 #12 0x0806ab4b in main ()
 #13 0x420158f7 in __libc_start_main () from /lib/i686/libc.so.6
 (gdb) 

Ok, this is a Win9x client scanning for print jobs. Up the
lpq cache time I would suggest. What Samba version ?

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


Re: [Samba] extremely high %CPU and TIME

2003-11-19 Thread Tom Ryan
2.2.7 on redhat 8. i uped lpq cache time = 60. doesn't seem to help.

Tom

On Wed, 19 Nov 2003, Jeremy Allison wrote:

 On Wed, Nov 19, 2003 at 03:23:31PM -0500, Tom Ryan wrote:
  Jeremy,
  
  (gdb) bt
  #0  0x08173176 in tdb_brlock ()
  #1  0x0817349a in tdb_unlock ()
  #2  0x0817517a in tdb_next_lock ()
  #3  0x081752a6 in tdb_traverse ()
  #4  0x08179de4 in print_queue_status ()
  #5  0x0807a6de in api_DosPrintQGetInfo ()
  #6  0x0807fdd8 in api_reply ()
  #7  0x08077e03 in reply_trans ()
  #8  0x080af1af in switch_message ()
  #9  0x080af3a9 in construct_reply ()
  #10 0x080af528 in process_smb ()
  #11 0x080b01b4 in smbd_process ()
  #12 0x0806ab4b in main ()
  #13 0x420158f7 in __libc_start_main () from /lib/i686/libc.so.6
  (gdb) 
 
 Ok, this is a Win9x client scanning for print jobs. Up the
 lpq cache time I would suggest. What Samba version ?
 
 Jeremy.
 

-- 
___
Tom RyanVoice: 856-225-6361
Consulting System Administrator   Fax: 856-969-7900
Rutgers School of Law - Camden

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


Re: [Samba] extremely high %CPU and TIME

2003-11-19 Thread Jeremy Allison
On Wed, Nov 19, 2003 at 03:36:15PM -0500, Tom Ryan wrote:
 2.2.7 on redhat 8. i uped lpq cache time = 60. doesn't seem to help.

How many print jobs do you have on this box ? Hmmm. As I recall,
a Win9x client with an open print monitor will just continually
scan the server - pounding it with requests. 

We redesigned this for 3.0.x (try 3.0.1pre3) but it won't fix
the problem that Win9x clients will continuously scan the print
server - this is a problem of the Win9x design, not of Samba.
The same thing happens to WinNT/2k/2k3.

Can you get people to close down the print monitors ?

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


Re: [Samba] extremely high %CPU and TIME

2003-11-19 Thread Tom Ryan
Jeremy,

All clients are windows xp.

Tom

On Wed, 19 Nov 2003, Jeremy Allison wrote:

 On Wed, Nov 19, 2003 at 03:36:15PM -0500, Tom Ryan wrote:
  2.2.7 on redhat 8. i uped lpq cache time = 60. doesn't seem to help.
 
 How many print jobs do you have on this box ? Hmmm. As I recall,
 a Win9x client with an open print monitor will just continually
 scan the server - pounding it with requests. 
 
 We redesigned this for 3.0.x (try 3.0.1pre3) but it won't fix
 the problem that Win9x clients will continuously scan the print
 server - this is a problem of the Win9x design, not of Samba.
 The same thing happens to WinNT/2k/2k3.
 
 Can you get people to close down the print monitors ?
 
 Jeremy.
 

-- 
___
Tom RyanVoice: 856-225-6361
Consulting System Administrator   Fax: 856-969-7900
Rutgers School of Law - Camden

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


Re: [Samba] extremely high %CPU and TIME

2003-11-19 Thread Jeremy Allison
On Wed, Nov 19, 2003 at 03:42:11PM -0500, Tom Ryan wrote:
 Jeremy,
 
 All clients are windows xp.

Ok, then you're not using the spoolss pipe code on the Samba
server, as this scanning behaviour goes away once WNT or above
clients can open the SPOOLSS pipe (they use a strange form of
change notify instead of scanning).

Do you have disable spoolss set in your smb.conf ?

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


Re: [Samba] extremely high %CPU and TIME

2003-11-19 Thread Tom Ryan
yes.. but I was having problems before with that enabled previously.

On Wed, 19 Nov 2003, Jeremy Allison wrote:

 On Wed, Nov 19, 2003 at 03:42:11PM -0500, Tom Ryan wrote:
  Jeremy,
  
  All clients are windows xp.
 
 Ok, then you're not using the spoolss pipe code on the Samba
 server, as this scanning behaviour goes away once WNT or above
 clients can open the SPOOLSS pipe (they use a strange form of
 change notify instead of scanning).
 
 Do you have disable spoolss set in your smb.conf ?
 
 Jeremy.
 

-- 
___
Tom RyanVoice: 856-225-6361
Consulting System Administrator   Fax: 856-969-7900
Rutgers School of Law - Camden

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


Re: [Samba] extremely high %CPU and TIME

2003-11-19 Thread Tom Ryan
I enabled that.. but I'm still having the same issue.

On Wed, 19 Nov 2003, Tom Ryan wrote:

 yes.. but I was having problems before with that enabled previously.
 
 On Wed, 19 Nov 2003, Jeremy Allison wrote:
 
  On Wed, Nov 19, 2003 at 03:42:11PM -0500, Tom Ryan wrote:
   Jeremy,
   
   All clients are windows xp.
  
  Ok, then you're not using the spoolss pipe code on the Samba
  server, as this scanning behaviour goes away once WNT or above
  clients can open the SPOOLSS pipe (they use a strange form of
  change notify instead of scanning).
  
  Do you have disable spoolss set in your smb.conf ?
  
  Jeremy.
  
 
 

-- 
___
Tom RyanVoice: 856-225-6361
Consulting System Administrator   Fax: 856-969-7900
Rutgers School of Law - Camden

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


Re: [Samba] extremely high %CPU and TIME

2003-11-19 Thread Jeremy Allison
On Wed, Nov 19, 2003 at 03:57:56PM -0500, Tom Ryan wrote:
 yes.. but I was having problems before with that enabled previously.

The client behaviour changes once spoolss is enabled. If you're 
claiming to see the same issue on the clients you need to restart
the spooler service on them in order for them to notice the now
open spoolss pipe.

3.0.x has a *much* improved version of the printing code we use
on the HP PSA (Print Server Appliance), so I can definately
recommend it.

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


Re: [Samba] extremely high %CPU and TIME

2003-11-19 Thread Tom Ryan
We are in the middle of planning our upgrade.

Thanks will try this and report back!

Tom

On Wed, 19 Nov 2003, Jeremy Allison wrote:

 On Wed, Nov 19, 2003 at 03:57:56PM -0500, Tom Ryan wrote:
  yes.. but I was having problems before with that enabled previously.
 
 The client behaviour changes once spoolss is enabled. If you're 
 claiming to see the same issue on the clients you need to restart
 the spooler service on them in order for them to notice the now
 open spoolss pipe.
 
 3.0.x has a *much* improved version of the printing code we use
 on the HP PSA (Print Server Appliance), so I can definately
 recommend it.
 
 Jeremy.
 

-- 
___
Tom RyanVoice: 856-225-6361
Consulting System Administrator   Fax: 856-969-7900
Rutgers School of Law - Camden

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


Re: [Samba] extremely high %CPU and TIME

2003-10-13 Thread Tom Ryan
Does anyone have any thoughts on this? Any things that I can try? I'm open 
to suggestions :)

Tom

On Thu, 9 Oct 2003, Tom Ryan wrote:

 Ok.. still researching this. should i try disabling kernel oplocks? also, 
 any thoughts on why it only happens to certain user processes?
 
 could it be a corrupted file or patch level of their machine?
 
 Tom
 
 On Thu, 9 Oct 2003, Tom Ryan wrote:
 
  Looking around on the net, I saw that someone else had a similar issue 
  when they were running all of their users behind terminal server. Our 
  users are all behind two NAT'd firewalls.
  
  Is it possible that there is some confusion going on? should I stop the 
  service and remove locking.tdb? (that's what 14 references right?)
  
  Thanks.
  
  tom
  
  On Wed, 8 Oct 2003, Tom Ryan wrote:
  
   2.4.20-19.8 on redhat 8
   
   strace reports 
   fcntl64(14, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=628, len=1}, 
   0xb150) = 0
   fcntl64(14, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=632, len=1}, 
   0xb150) = 0
   
   over and over again (with different start points)
   
   Tom
   
   On Wed, 8 Oct 2003, Jeremy Allison wrote:
   
On Wed, Oct 08, 2003 at 02:29:31PM -0400, Tom Ryan wrote:
 I have a samba box that has around 70 users on it at any given time.
 
 When I run ps axu there are certain smbd processes (and usually always 
 by the same login) that look like
 
 username 26126 39.4  0.1  6652 2912 ?S13:44  17:12 smbd -D
 
 any ideas why CPU is running at 39.4??

What OS ? Try doing an strace on it to see what the process is doing.

Jeremy.

   
   
  
  
 
 

-- 
___
Tom RyanVoice: 856-225-6361
Consulting System Administrator   Fax: 856-969-7900
Rutgers School of Law - Camden

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


Re: [Samba] extremely high %CPU and TIME

2003-10-09 Thread Tom Ryan
Looking around on the net, I saw that someone else had a similar issue 
when they were running all of their users behind terminal server. Our 
users are all behind two NAT'd firewalls.

Is it possible that there is some confusion going on? should I stop the 
service and remove locking.tdb? (that's what 14 references right?)

Thanks.

tom

On Wed, 8 Oct 2003, Tom Ryan wrote:

 2.4.20-19.8 on redhat 8
 
 strace reports 
 fcntl64(14, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=628, len=1}, 
 0xb150) = 0
 fcntl64(14, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=632, len=1}, 
 0xb150) = 0
 
 over and over again (with different start points)
 
 Tom
 
 On Wed, 8 Oct 2003, Jeremy Allison wrote:
 
  On Wed, Oct 08, 2003 at 02:29:31PM -0400, Tom Ryan wrote:
   I have a samba box that has around 70 users on it at any given time.
   
   When I run ps axu there are certain smbd processes (and usually always 
   by the same login) that look like
   
   username 26126 39.4  0.1  6652 2912 ?S13:44  17:12 smbd -D
   
   any ideas why CPU is running at 39.4??
  
  What OS ? Try doing an strace on it to see what the process is doing.
  
  Jeremy.
  
 
 

-- 
___
Tom RyanVoice: 856-225-6361
Consulting System Administrator   Fax: 856-969-7900
Rutgers School of Law - Camden

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


Re: [Samba] extremely high %CPU and TIME

2003-10-08 Thread Tom Ryan
Jeremy,

2.4.20-19.8 on redhat 8

strace reports 
fcntl64(14, F_SETLKW64, {type=F_UNLCK, whence=SEEK_SET, start=628, len=1}, 
0xb150) = 0
fcntl64(14, F_SETLKW64, {type=F_WRLCK, whence=SEEK_SET, start=632, len=1}, 
0xb150) = 0

over and over again (with different start points)

Tom

On Wed, 8 Oct 2003, Jeremy Allison wrote:

 On Wed, Oct 08, 2003 at 02:29:31PM -0400, Tom Ryan wrote:
  I have a samba box that has around 70 users on it at any given time.
  
  When I run ps axu there are certain smbd processes (and usually always 
  by the same login) that look like
  
  username 26126 39.4  0.1  6652 2912 ?S13:44  17:12 smbd -D
  
  any ideas why CPU is running at 39.4??
 
 What OS ? Try doing an strace on it to see what the process is doing.
 
 Jeremy.
 

-- 
___
Tom RyanVoice: 856-225-6361
Consulting System Administrator   Fax: 856-969-7900
Rutgers School of Law - Camden

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