Efficient monitoring of CRAB jobs at CMS

Abstract

CRAB is a tool used for distributed analysis of CMS data. Users can submit sets of jobs with similar requirements (tasks) with a single request. CRAB uses a client-server architecture, where a lightweight client, a server, and ancillary services work together and are maintained by CMS operators at CERN. As with most complex software, good monitoring tools are crucial for efficient use and longterm maintainability. This work gives an overview of the monitoring tools developed to ensure the CRAB server and infrastructure are functional, help operators debug user problems, and minimize overhead and operating cost. This work also illustrates the design choices and gives a report on our experience with the tools we developed and the external ones we used.