Print

Print


  Hi Jacek,

On Tue, Sep 14, 2004 at 09:59:25AM -0700, Jacek Becla wrote:
> > server side monitoring
> 
> - found bug in receiver & decoder code (50% of each trace packet 
> ignored), easy to fix.
> - loaded data from recent 1-server test to mysql, if I adjust the 
> numbers to account for the bug, average 88MB/file read, largest chunk 
> read 3.2MB, have many more statistics.

  Can you break up the statistics by "PR", "SP", "PRskims" and "SPskims"?
(I suspect that they should have different distributions...)

> - recent test proved that the server-side code is non-intrusive. We 
> could try to turn monitoring on all kan servers (once I fix the bug in 
> receiver). How about next week? I have a dedicated machine for 
> collecting data (objydev11) and enough nfs space to accommodate 
> collector's log files.
> - per recent email exchange: hn for access patterns related discussion: 
> KanSOS

  For the detailed discussions BaBar specific stuff (which most of it will be 
initially). Occasional summaries of how the system works and the information 
it provides might be interesting for the others here, though.

> - thought more about issue discussed last week: parsing servers' log 
> files. I second Andy's opinion about duplicating efforts, could _easily_ 
> do that via code I already have

  But not so easily at all sites, which is what I want. It is much easier
for people to cough up their log files than to setup the full monitoring
system with collector and relational database (and then synch that). We'll 
try that just at SLAC as it will provide more information than we need for 
the immediate use case. (In fact we can 2/3 of the way there just collecting
the redirector log files with the "odc.trace redirect" turned on.)

> > administrative interface
> 
> - I've started looking at xrdclient code to familiarize with xrd code. 
> Goal to implement 2 simple admin commands (e.g. mv, rm) in the next 2-3 
> weeks. At the same time, discuss list of admin commands that need to be 
> implemented and prioritize (did not have time to study discussion about 
> data mgmt cases sent around earlier)

  Ok, I'd like to discuss the admin interface with you, Andy and Fabrizio,
though. When are you back from FNAL?

                                   Pete

-------------------------------------------------------------------------
Peter Elmer     E-mail: [log in to unmask]      Phone: +41 (22) 767-4644
Address: CERN Division PPE, Bat. 32 2C-14, CH-1211 Geneva 23, Switzerland
-------------------------------------------------------------------------