 |
 |
Hi Everyone, Recently, there has been some discussion and feedback concerning the hourly lightweight job submission tests. The test jobs are sumitted to a site directly (globus-job-run), and through each of the resource brokers (edg-job-submit). With the increase in LCG resources observed over the last few months, especially in the number of RBs, the total number of jobs sent to each site has increased from a "few-an-hour" to more than 10 per hour. The purpose of these test jobs is to identify when critical middleware components fail e.g. the gatekeepers and RBs. In general, site admins have observed cases where there are many jobs piling up at sites either due to limited worker node resoucres, or other problems e.g. the RB gridftp problems reported here last week. Clearly, monitoring the grid shouldn't be an intrusive procedure, although there are those that would argue that high job submission rates are necessary for the robust testing of middleware components etc. As a result, we have disabled job submissions through all the RBs, except the testzone RB at CERN. Dave ========================================================= Dr Dave Kant CCLRC eScience Department Phone: (+44)|(0) 1235 778178 Rutherford Appleton Laboratory Fax: (+44)|(0) 1235 446626 Chilton, Didcot, Oxon, OX11 0QX, UK Email: [log in to unmask] ==========================================================