poor Samba performance

[Erich Z]

Hello
following situtation:
Samba works well for us for many years. Now we found it tooks too long to copy a new fileset over an existing fileset on our server (HP-UX 11.0 L1000).
Our Sample:
2,6GB - 77705 Files within 2443 Folders
My copy test:
xcopy /S /E /I /Y S:\home\programme\v5\B14\intel_a D:\tmp\cpy\verz
From Server (Samba) to XPworkstation (my PC):
30 Min for copy and 30 Min for copy over existing
Same Values with copy from my workstaion to a other windows (XP) box
Same copy to a samba share:
150 Min !!!! (4 tests 151,205,148,150)
Have made tests with our server and a HP C3750 Workstation (because ist easyser to test).
Samba config attached. Nobody is working on this Workstation.

Have tested it with samba 2.2.a and now upgraded to 2.2.l (same thing).

Have found following with google:
what is the bottleneck ?? CPU, MEM, Network, DISK, swap space??
And thats my question have checked this things everthing seams ok:
top:
System: ecshp033 Wed Feb 9 16:56:34 2005
Load averages: 1.60, 1.56, 1.37
104 processes: 100 sleeping, 4 running
Cpu states:
LOAD USER NICE SYS IDLE BLOCK SWAIT INTR SSYS
1.60 84.0% 0.0% 2.4% 13.6% 0.0% 0.0% 0.0% 0.0%

Memory: 94916K (37888K) real, 83104K (37052K) virtual, 1479996K free Page# 1/5

TTY PID USERNAME PRI NI SIZE RES STATE TIME %WCPU %CPU COMMAND
? 23817 root 233 20 12420K 7936K run 17:06 84.12 83.97 smbd
? 31 root 152 20 4864K 0K run 3:59 0.61 0.61 vxfsd
? 1204 daemon 154 20 62040K 14980K sleep 108:46 0.32 0.32 X
? 23803 root 154 20 3992K 868K sleep 0:00 0.21 0.21 nmbd
# swapinfo -t
Kb Kb Kb PCT START/ Kb
TYPE AVAIL USED FREE USED LIMIT RESERVE PRI NAME
dev 4194304 0 4194304 0% 0 - 1 /dev/vg00/lvol2
reserve - 137484 -137484
memory 1460532 137848 1322684 9%
total 5654836 275332 5379504 5% - 0 -

# sar -d 1 10
HP-UX ecshp033 B.11.00 U 9000/785 02/09/05
Average c3t6d0 9.49 0.50 26 406 3.91 5.53

Network: nobody is working there (and read is fast).
My conclusion: Samba is slow

Is that so or
What can I do?? Settings?
Any help welcome thx in advance
Erich

[B. Hulst]

Hi,

It doesn't look like you have a memory or cpu problem.

Has anything changed in the network?

What is the FTP speed to and from the server?

Regards,
Bob

[Erich Z]

Hello,
Sorry have forgotten. Normal 100Mbit performance in our Network. Have made the same test with a 217M file (takes 1:10 in all directions => sampletime should be about 14 min - the overhead to 30 does not matter a real problem are the two extra hours).

I'm not thinking its a direct networkproblem and its not depending on overwriting also not on the used hardware.

I think the troubles source is the handling of many smal files with samba - and I am hoping there is a positve setting.

mfg Erich



P.S. I think within this group are the samba exprerts - for me samba and Cifs is the same

[Geoff Wild]

One setting I use that improved performance in smb.conf:

keepalive = 0

Are you running winbind?

I have seen performance issues when the samba share is also a NFS share....

Rgds...Geoff

[Erich Z]

Hello,
Have checked keepalive - seems to be no difference because set on our server (with socket option so_keepalive) and not set as you have seen on our testbox.

winbind is not running.

NFS does not matter because on the server there are many NFS shares shares (but these are submounts to the Windows share) on my testworkstation there are only NFS mounts but no exported shares.

mfg Erich

[Steven E. Protter]

You might want to measure performance to see what the real problem is.

copying a large number of files can cause kernel issues. I'm attaching some perf monitor scripts, given to me by HP Support(no warranty) and made to work correctly in the background by me(again no warranty). They should help you identify the problem.

Also, a network card getting ready to fail will sometimes have slowdowns. You would notice this on logins and other network transfers.

Also a problem with Fiber cards/patching and or the vxfs filesysetm i/o hang would impact Samba peformance.

Yeah, its a fishing expedition. Sorry about that. Consider these two patches if relavent to your system.

From a current support call I have open after an i/o hang. rp5450 system.

PHCO_29379 1.0 LVM commands cumulative patch
PHKL_26743 1.0 LVM Cumulative Patch
PHKL_28984 1.0 Fibre Channel Mass Storage Patch
PHKL_29365 1.0 SCSI IO Cumulative Patch
PHKL_29539 1.0 PDC_SCSI_PARMS,iCOD,vPars,IPMI,OLA/R
PHKL_29685 1.0 SCSI Ultra160 Cumulative Patch
PHKL_25506 1.0 asyncio driver patch

Except for the FCMS and asyncio driver patches installed, there are
later HP recommended versions for all of the above.

Recommend updating system with the following patches (plus any patch
dependencies needed with these):

PHSS_28509 s700_800 11.11 Tachyon TL Fibre Channel Driver Patch
PHKL_32235 s700_800 11.11 VxFS cumulative patch for three way deadlock
PHCO_30698 s700_800 11.11 LVM commands cumulative patch; LVM OLR
PHKL_31216 s700_800 11.11 LVM Cumulative Patch; LVM OLR; SLVM 16 Node
PHKL_30510 s700_800 11.11 SCSI IO Cumulative Patch
PHKL_30523 s700_800 11.11 PDC_SCSI_PARMS,iCOD,vPars,IPMI,OLA/R
PHKL_30511 s700_800 11.11 SCSI Ultra160 Cumulative Patch


SEP