FBSNG v1.5a Release Notes ------------------------- This document summarizes the changes made in FBSNG v1.5 comparing to v1.4 and describes the upgrade procedure. What in new in FBSNG v1.5a ? ---------------------------- New Features: ------------- * Relative computer power is introduced as new node class attribute "power". This attribute is used for more accurate fair-share scheduling and for time limit adjustment on per-node basis. * Minimal NICE level for process type is introduced. Processes of the process type are executed with nice level determined as sum of one defined for the process type and specified by the user in the job description. * -C option added to "fbs exec". If the job does not start within time specified with -w, it will be cancelled. * Added support for Globus Toolkit 2.2 interface Major Bugs Fixed: ----------------- * Optional configuration parameter sets the limit for the time spent by the Scheduler in single uninterrupted run. This option makes sure that even when the Scheduler is busy, clients still can get response from the Batch Manager in reasonably short time between Scheduler runs. * Sections of different users in the same queue are scheduled in round-robin fasion. * Inter-section dependency evaluation algorithm was optimized to avoid long delays in case of very large jobs (~300 dependent sections). * FBSWWW communication with the client is made asynchronous to avoid service lock-up by a malicious or inoperative HTTP client or Internet security scanning software. FBSWWW now spawns separate subprocess to handle each individual client. Number of concurrently running FBSWWW clients is limited by configuration parameter. * "Wildcard" kill operation is made interruptable by ^C. * -1 option added to "fbs submit" command. If specified, the submit prints only new job id without any words. * When batch processes are failing at an excessive rate on a farm node, the node will be put on hold automatically. This will help avoid situations when one or more batch queues are "drained" through a node. * -u argument of ls, lj and status subcommands will use text Username field instead of trying to convert it to numeric UID. * Recovery procedure was handling incorrectly the case when bmgr would be shut down while starting a batch process (fixed after release of v1.5) * Eliminated excessive "section exceeded time limit" messages in section log (fixed after v1.5) Upgrading FBSNG to v1.5a in FUE environment ------------------------------------------- 1. Hold all FBSNG queues (fbs hold queue all). 2. Shut down all FBSNG daemons using $FBSNG_ROOT/bin/shutdown_*.sh scripts. To shut down launchers, you must log in as root. 3. If necessary, install or upgrade FCSLIB to v2_1a; 4. If necessary, as user "products", install and declare desired version of Python. 5. As user "products", install, declare and tailor FBSNG v1.5 for all required platforms. If you are using UPS to install fbsng, use "upd install -R ..." command. This will install FCSLIB too. Optionally, declare fbsng with "local_python" qualifier (ups declare ... -q local_python ...) for the patform(s) farm nodes are running. See FBSNG Installation and Administration Guide for tailoring instructions. 6. If Kerberos support is required, install and declare Kerberos for all platforms following Kerberos installation instructions; 7. Initialize FBSNG v1.5 environment (issue setup fbsng). 10. If you are going to run FBSWWW server, add FBSWWW configuration information to fbs.cfg. For example: %set fbswww port = 8080 graphs_dir = /tmp/fbswww_graphs Copy start_fbswww.sh, kill_fbswww.sh, shutdown_fbswww.sh from $FBSNG_DIR/templates/bin to $FBSNG_ROOT/bin. Use these scripts to start, restart and shut down FBSWWW server. 11.Run "fbs chkcfg" to validate the configuration. 12.Start FBSNG v1.5 components using $FBSNG_ROOT/bin/start_*.sh scripts. To start launchers, you must log in to farm nodes as root. 13.Remove history.run.cron script from cron table as it is obsolete. For more details on FBSNG configuration and shutting down and starting FBSNG components, see FBSNG Installation and Administration Guide.