Hi,

On Wed, 18 Feb 2015, Marcus Ebert wrote:

> 2) xprep seems to be the tool which could be used for this purpose. On all 
> file servers automatic staging in from old-cluster works on demand through 
> frm_xfrd, but when using xprep nothing is staged in. I used
> xprep -s -f FileList.txt metamanger
> and I see then in the cmsd metamanger log file messages about SelPrep for all 
> files failed; No servers are available to read the file
>
Ok, for that I found out it needs to run a prep manager, but haven't found 
any documentation about it so far.

>
> Could you please let me know how you handle such situations for your clusters 
> and how xprep should work (probably I misunderstand right now the use case 
> for it).
>
I found the tool "xrdstagetool" which seems to do what I want.
However, there are again 2 problems with that. First let me show you the 
configuration I have so far which isn't any complicated I think:
- 1 Metamanger at a Computing Center (GridKa)
- 1 Redirector and 4 servers at GridKa, servers subscribe only to this 
redirector which subscribes to the metamanager
- 1 redirector and 2 server at a European university, server subscribe 
only to the local redirector which subscribes to the metamanager
- 1 redirector and 1 server at Canadian university, again server only 
subscribes to the local redirector which subscribes to the metamanager

What is working now is to ask the Metamanger for any file on any of the 
servers and I can get it. It also works to send a file to the Metamanger 
using xrdcp and it will end up on one of the servers.


However, with xrdstagetool there are 2 problems:
1) If I connect xrdstagetool to the metamaner for staging, then the 
metamanger asks all redirectors/servers if anyone has the file. Since it 
should stage in from somewhere else, no server has it so far. The 
Metamanger replies to xrdstagetool that no server are available to read 
the file.
How is it possible that in that case that no server has the file, the 
metamanger triggers the staging process on one of the servers?

2) If I connect xrdstagetool directly to one of the redirectors then it 
works. The servers don't have the file and on one of them the staging is 
initiated. Once the file is local, the answer is send to xrdstagetool 
which finish then successfull.
However, this doesn't work with xrootd-4.1.1. With that version there is 
no notification about a successfull staging if the staging process takes 
too long (>5min). The staging of the file is still successfull but 
xrdstagetool never receives an answer and times out after a while when the 
staging is already long finished.
I see in the xrd log file when the staging starts that the client should 
wait:
XrootdProtocol: 0100 delaying client up to 1800 sec
and then normally something like:
ofs_recvEvent: -->stage OK
But this last notification doesn't appear with xrootd-4.1.1 if the 
transfer takes long, using xrootd-v4.0.4 it works.
Is this a known issue?


Could anyone which uses a metamanger please let me know how you have 
configured that xrdstagetool is working?


Thanks,
   Marcus
-- 
--------------------------------------------------------------------
Marcus Ebert
SLAC National Accelerator Laboratory
Mail Stop 34
2575 Sand Hill Road
Menlo Park, CA, 94025
USA
------
office: ROB 129
Tel.  : +1 (650) 926-3565
====================================================================
      Please, don't send Word, Excel or PowerPoint attachments -
            use pdf, plain text or HTML format instead.
--------------------------------------------------------------------

########################################################################
Use REPLY-ALL to reply to list

To unsubscribe from the XROOTD-L list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-L&A=1