ADSM-L

Restore problem with big filesystem

2005-02-02 04:24:57
Subject: Restore problem with big filesystem
From: Rainer Wolf <rainer.wolf AT KIZ.UNI-ULM DOT DE>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Wed, 2 Feb 2005 10:20:25 +0100
Hello,
has someone successfully restored a filesystem with more than 3 mio files ?
Currently our 'restore' is now running 20 hours and it still shows

ANS1899I ***** Examined 2,689,000 files *****
ANS1899I ***** Examined 2,690,000 files *****
ANS1899I ***** Examined 2,691,000 files *****
...
Because there are about 3.5 mio files to be restored it may run for a
very long time. At this stage there is no byte 'restored' yet  --- just
doing this 'Examination'.

The dsmc on the solaris(10) client-system ( v440, 4 cpus, 16GB mem ) is showing

CPU    PID User      NI  State   Size    RSS  CPU% SCPU%   CPU-Time Command
  4T  8970 root       0     on  1011m  1009m 100.4   0.0   12:30:48 dsmc

... there is nothing else running there.

I seems that the 'restore' slows down with the number of files 'examined'

We have both tried V5.3.0.0 solaris- client and  this current session
has client Version 5.1.6.5 .
The process is excessively doing string-operation like 'strcmp' and 'strlen'.

The restore - command is:
dsmc restore -quiet -virtualnode=mail /mail/ /data2/mail/

The /data2 -fs is empty and our dsm.sys/dsm.opt looks like:

blackhole:...//# cat /opt/tivoli/tsm/client/ba/bin/dsm.opt
subdir  yes
testflag disablenqr
blackhole:...//# cat /opt/tivoli/tsm/client/ba/bin/dsm.sys
servername         adsmaix
commmethod         tcpip
tcpport            1500
tcpserveraddress   xxxxxxx
passwordaccess     generate
schedlogname       /var/adm/dsmsched.log
errorlogname       /var/adm/dsmerror.log
schedlogretention  7
errorlogretention  14
inclexcl           /opt/tivoli/tsm/client/ba/bin/exclude.lis
tcpnodelay         yes
blackhole:...//#

The 'testflag disablenqr' we used because at least for a smaller number of
files ( eG 10000 files ) it had a tremendeous performance effect.

The server is showing often SendWait - currently up to 20 s WaitTime:

  Sess    Comm.     Sess        Wait      Bytes      Bytes    Sess     Platform 
   Client Name
Number    Method    State       Time       Sent      Recvd    Type
------    ------    ------    ------    -------    -------    -----    -------- 
   --------------------
26,610    Tcp/Ip    SendW      20 S     689.8 M        496    Node     SUN SOL- 
   MAIL
                                                                        ARIS


Anyone knows whats going on here ?

Greetings,
Rainer



--
--------------------------------------------------------------------------
Rainer Wolf                              Mail:  rainer.wolf AT kiz.uni-ulm DOT 
de
Kommunikations und Informationszentrum   Tel/Fax:  ++49 731 50-22482/22471
Abt. Infrastruktur, Uni Ulm              Web:       http://kiz.uni-ulm.de/

<Prev in Thread] Current Thread [Next in Thread>