[jcifs] Reading a directory tree

[Thomas Bley]

Hi,

I've seen this when doing load tests with the webdisk (jcifs 1.2.1 and 
1.1.9).
The fact is that the filer is not overloaded, but suddenly doesn't get 
any more requests from jcifs. When I got this error message, it was no 
longer possible to do any cifs-connection and I restarted tomcat.

I've used httperf to test the webdisk: doint 90 http-requests per second 
for a time of 10 seconds (the webserver was a dual xeon with 2 GB ram 
and SuSE Enterprise Linux and Tomcat 5.0.x, the fileserver a netapp machine)

I got less problems when I used a smaller thread stacksize of 256k, e.g.
CATALINA_OPTS = "-Xms1536m -Xmx1536m -Xss256k"

maybe this has any dependancy ? (the maximum ram usage I got was 1gb)


The command I've used for httperf:

httperf --hog --num-conn 900 --rate 90 --think-timeout=30 
--http-version=1.1 --server 192.168.1.2 --port=8080 
--uri=/ba/base/192.168.1.1/myuser/download_test/?username=myuser\&password=mypass\&domain=mydomain\&server=192.168.1.1\&login=Login


This request mainly does this:

ntlm = new NtlmPasswordAuthentication(...)
SmbFile pathDirs = new SmbFile( "smb://192.168.1.1/", ntlm );
SmbFile[] dirs = pathDirs.listFiles();

for (int i=0; i<dirs.length; i++) {
   out.println(...);
}

SmbFile pathDirs = new SmbFile( "smb://192.168.1.1/myuser/", ntlm );
SmbFile[] dirs = pathDirs.listFiles();

for (int i=0; i<dirs.length; i++) {
   out.println(...);
}

bye
Tom


Michael B Allen wrote:
> On Fri, 16 Sep 2005 09:40:03 +0100
> "Mike Streeton" <mike.streeton at ardentia.co.uk> wrote:
> 
> 
>>I am reading a directory using JCIFS, which works okay for about 15
>>minutes 
>>and the starts producing:
>>com.ardentia.netsearch.exceptions.FileException: jcifs.smb.SmbException:
>>jcifs.util.transport.TransportException: 
>>Transport1[fileserver1<20>/192.168.20.1:445] timedout waiting for
>>response 
>>to 
>>Trans2QueryPathInformation[command=SMB_COM_TRANSACTION2,received=false,e
> 
> 
> Other people have reported similar problems. There could be a concurrency
> flaw in the new transport. Try 1.1.11.
> 
> Mike
>