Possible clue to samba/explorer performance problem

Ron Alexander rcalex at home.com
Wed Nov 29 13:03:58 GMT 2000


I have (and I am sure many/all of you have also) seen complaints of slow
samba performance posted many times before. I thought I would forward this
one to the group as it might offer a clue.

What I find odd is that the performance was good and only became not good
after re-starting the samba daemons.

Does anyone have an idea as to what is happening here?

Regards,
Ron Alexander (Stratus VOS Samba 2 support)

"Lyall Pearce" <lyallp at nospam.bigpond.com> wrote in message
news:<3A24B6CB.D18AD8CB at nospam.bigpond.com>...
> I tried what you suggested Ron - here is what I did.
>
> 1. Powered up the Win machine (the linux machine has been up for days)
> 2. Logged on the win machine
> 3. tried copying an 11mb file using explorer  (both ways, from and to the
linux
> machine, multiple times)- bang, 11mb copied in no time flat! - Damn (well
good,
> if it would only stay this way!)
> 4. stopped and restarted both Samba daemons using swat on the linux
machine.
> 5. tried copying the 11mb file - SLOW write
> 6. tried the same thing using dos command prompt and COPY - SLOW write
> 7. Rebooted the Win machine
> 8. Restarted the Samba daemons while the win machine was in POST
> 9. Logged in
> 10. Tried the copy using explorer - SLOW (damn, now I am really confused)
> 11. Turned on auto refresh in swat.
> 12. Tried the copy again - swat looks like :-
>
> Active Connections
>
>    PID        Client             IP address
> Date                                        Kill
>    27971       trans-pc      192.168.0.2                   Wed Nov 29
18:16:09
> 2000
>
>   Active Shares
>
>    Share        User           Group                 PID
> Client                                  Date
>    lyallp       lyallp           lyallp
> 27971                     trans-pc                           Wed Nov 29
> 18:16:09 2000
>
>
>   Open Files
>
>    PID         Sharing                  R/W
> Oplock
> File
Date
>    27971       DENY_ALL       WRONLY                 EXCLUSIVE+BATCH
> /home/lyallp/temp/BIG.DAT               Wed Nov 29 18:22:41 2000
>
> It all seems quite ok to my uneducated eye.
>
> Suggestions?
>
> Yours hopefully
>
> ....Lyall
>
> Ron Alexander wrote:
>
> > Try another test using the dos copy command from a dos box.
> >
> > There are known problems(msft) with explorer.
> >
> > "Lyall Pearce" <lyallp at nospam.bigpond.com> wrote in message
> > news:20001128.10554300 at lyalls-pc.home.net.au...
> > Further information regarding my problem.
> >
> > Linux Redhat 6.2ish server
> > Win 98 2nd ed
> >
> > My copy technique is to 'copy' a 11mb file using explorer on the Win
> > machine and paste it into my mounted home directory (the linux machine).
> >
> > I tried a copy of the file - slow write
> > Stop both smb and nmd, then re-start them using swat. Slow write still.
> > I performed an FTP from the windows machine to the server on the same
> > 11mb file. 3Mb per second transfer rate for both read and write.
> > I did a tcpdump on the interface while using the explorer cut/paste
> > method. I started the trace, went to the windows machine and pressed
> > CTRL-V, waited 15 seconds then aborted the copy on the windows machine,
> > then terminated the tcpdump.
> > I captured the packets from the tcpdump in a file for later analysis.
> >
> > If anyone wants to look at this dump, let me know.
> >
> > Now, I am no expert at all when it comes to TCP/IP packets but looking
at
> > the packet timestamps, there appear to be regular multi-second 'gaps'...
> > Like this...
> >
> > 21:07:20.965543 < trans-pc.home.net.au.1098 >
> > lyalls-pc.home.net.au.netbios-ssn: P 132601:132706(105) ack 1589 win
8470>>>
> > NBT (DF) (ttl 128, id 12300)
> > 21:07:20.965668 > lyalls-pc.home.net.au.netbios-ssn >
> > trans-pc.home.net.au.1098: P 1589:1628(39) ack 132706 win 32120>>> NBT
(DF)
> > (ttl 64, id 56314)
> > 21:07:20.965912 < trans-pc.home.net.au.1098 >
> > lyalls-pc.home.net.au.netbios-ssn: P 132706:132811(105) ack 1628 win
8431>>>
> > NBT (DF) (ttl 128, id 12556)
> > 21:07:20.966030 > lyalls-pc.home.net.au.netbios-ssn >
> > trans-pc.home.net.au.1098: P 1628:1667(39) ack 132811 win 32120>>> NBT
(DF)
> > (ttl 64, id 56315)
> > 21:07:20.966287 < trans-pc.home.net.au.1098 >
> > lyalls-pc.home.net.au.netbios-ssn: P 132811:132916(105) ack 1667 win
8392>>>
> > NBT (DF) (ttl 128, id 12812)
> > 21:07:20.966405 > lyalls-pc.home.net.au.netbios-ssn >
> > trans-pc.home.net.au.1098: P 1667:1706(39) ack 132916 win 32120>>> NBT
(DF)
> > (ttl 64, id 56316)
> > 21:07:21.121147 < trans-pc.home.net.au.1098 >
> > lyalls-pc.home.net.au.netbios-ssn: . 132916:132916(0) ack 1706 win 8353
> > (DF) (ttl 128, id 13068)
> > 21:07:23.085261 < trans-pc.home.net.au.1098 >
> > lyalls-pc.home.net.au.netbios-ssn: P 132916:132984(68) ack 1706 win
8353>>>
> > NBT (DF) (ttl 128, id 13324)
> > 21:07:23.085429 > lyalls-pc.home.net.au.netbios-ssn >
> > trans-pc.home.net.au.1098: P 1706:1747(41) ack 132984 win 32120>>> NBT
(DF)
> > (ttl 64, id 56317)
> > 21:07:23.094627 < trans-pc.home.net.au.1098 >
> > lyalls-pc.home.net.au.netbios-ssn: . 132984:134444(1460) ack 1747 win
> > 8312>>>
> > NBT (DF) (ttl 128, id 13580)
> > 21:07:23.094748 < trans-pc.home.net.au.1098 >
> > lyalls-pc.home.net.au.netbios-ssn: . 134444:135904(1460) ack 1747 win
> > 8312>>>
> > NBT (DF) (ttl 128, id 13836)
> > 21:07:23.094773 > lyalls-pc.home.net.au.netbios-ssn >
> > trans-pc.home.net.au.1098: . 1747:1747(0) ack 135904 win 30660 (DF) (ttl
64,
> > id 56318)
> >
> > Every other packet going back and forth seems to be nice and snappy
> > judging by the decimal places after the second. The gaps seem to occurr
> > at no particular time interval or number of packets.
> >
> > Help??
>





More information about the samba-technical mailing list