=============================== Release Notes for PEGASUS 3.0 =============================== This is a major release that attempts to simplify configuration and running workflows through Pegasus. Existing users please refer to the Migration Guide available at http://pegasus.isi.edu/wms/docs/3.0/Migration_From_Pegasus_2_x.php . A user guide is now available online at http://pegasus.isi.edu/wms/docs/3.0/index.php NEW FEATURES -------------- 1) Support for new DAX Schema Version 3.2 Pegasus 3.0 by default parses DAX documents conforming to the DAX Schema 3.2 and is explained online at http://pegasus.isi.edu/wms/docs/3.0/api.php . The same link has documentation about using the JAVA/Python and Perl API's to generate DAX3.2 compatible DAX'es. Users are encourage to use the API's in their DAX Generators. 2) New Multiline Text Format for Transformation Catalog The default format for Transformation Catalog is now Text which is a multiline textual format. It is explained in the Catalogs Chapter http://pegasus.isi.edu/wms/docs/3.0/catalogs.php#transformation 3) Shell Code Generator Pegasus now has support for a Shell Code Generator that generates a shell script in the submit directory instead of Condor DAGMan and condor submit files. In this mode, all the jobs are run locally on the submit host. To use this set - pegasus.code.generator Shell - make sure that --sites option passed to pegasus only has site local mentioned. 4) Profiles and Properties Simplification Starting with Pegasus 3.0 all profiles can be specified in the properties file. Profiles specified in the properties file have the lowest priority. As a result of this a lot of existing Pegasus Properties were replaced by profiles. An exhaustive list of Properties replaced can be found at http://pegasus.isi.edu/wms/docs/3.0/Migration_From_Pegasus_2_x.php#id2765439 All Profile Keys are documented in the Profiles Chapter http://pegasus.isi.edu/wms/docs/3.0/advanced_concepts_profiles.php The properties guide was broken into two parts and can be found in the installation in the etc directory - basic.properties lists only the basic properties that need to be set to use Pegasus. Sufficient for most users. - advanced.properties lists all the properties that can be used to configure Pegasus. The properties documentation can be found online at http://pegasus.isi.edu/wms/docs/3.0/configuration.php http://pegasus.isi.edu/wms/docs/3.0/advanced_concepts_properties.php 5) Transfers Simplification Pegasus 3.0 has a new default transfer client pegasus-transfer that is invoked by default for first level and second level staging. The pegasus-transfer client is a python based wrapper around various transfer clients like globus-url-copy, lcg-copy, wget, cp, ln . pegasus-transfer looks at source and destination url and figures out automatically which underlying client to use. pegasus-transfer is distributed with the PEGASUS and can be found in the bin subdirectory . Also, the Bundle Transfer refiner has been made the default for pegasus 3.0. Most of the users no longer need to set any transfer related properties. The names of the profiles keys that control the Bundle Transfers have been changed . To control the clustering granularity of stagein transfer jobs following Pegasus Profile Keys can be used - stagein.clusters - stagein.local.clusters - stagein.remote.clusters To control the clustering granularity of stageout transfer jobs - stageout.clusters - stageout.local.clusters - stageout.remote.clusters 6) New tools called pegasus-statistics and pegasus-plots There are new tools called pegasus-statistics and pegasus-plots that can be used to generate statistics and plots about a worklfow run. The tools are documented in Monitoring and Debugging Chapter http://pegasus.isi.edu/wms/docs/3.0/monitoring_debugging_stats.php#id3083376 7) Example Workflows Pegasus distribution comes with canned examples that users can use to run after installing Pegasus. The examples are documented in the Example Workflows Chapter http://pegasus.isi.edu/wms/docs/3.0/example_workflows.php 8) Support for GT5 Pegasus now has support for GT5. Users can use GT5 only if they are using the new site catalog schema ( XML3 ) , as only that has a place holder to specify grid type with the grid gateways element. e.g <site handle="isi_viz" arch="x86" os="LINUX" osrelease="" osversion="" glibc=""> <grid type="gt5" contact="viz-login.isi.edu/jobmanager-pbs" scheduler="PBS" jobtype="compute"/> <grid type="gt2" contact="viz-login.isi.edu/jobmanager-fork" scheduler="Fork" jobtype="auxillary"/> .... </site> Users can use pegasus-sc-converter to convert their site catalogs to the XML3 format. Sample Usage: sc-client -I XML -O XML3 -i sites.xml -o sites.xml3 9) pegasus-analyzer can parse kickstart outputs pegasus-analyzer now parses kickstart outputs and prints meaningful information about the failed jobs. The following information is printed for the failed jobs - the executable - the arguments - the site the job ran on - remote the working directory of the job - exitcode - node the job ran on - stdout | stderr 11) Logging improvements pegasus-plan now has a -q option to decrease the logging verbosity. The logging verbosity can be increased by using the -v options. Additionally, there are two new logging levels - CONSOLE ( enabled by default - messages like pegasus-run invocation are printed to this level ) - TRACE ( more verbose that DEBUG level ). The INFO level is no longer enabled by default. To see the INFO log level messages pass -v to pegasus-plan To turn on the TRACE level pass -vvvv . In the TRACE mode, in case of exceptions the full stack trace for exceptions is printed. 12) Default condor priorities to jobs Pegasus assigns default Condor priorities to jobs now. The priorities come into play when running jobs in Condor vanilla / standard or local universe jobs. They dont apply for grid universe jobs. The default priority for various types of jobs is Cleanup : 1000 Stage out : 900 Dirmanager : 800 Stage in : 700 Compute jobs: level * 10 where level is the level of the job in the workflow as compute from the root of the workflow. This priority assignment gives priority to workflows that are further along, but also overlap data staging and computation (data jobs have higher priority, but the assumptions is that there are relatively few data staging jobs compared to compute jobs) . 13) Turning off Condor Log Symlinking to /tmp By default pegasus has the Condor common log -0.log in the submit file as a symlink to a location in /tmp . This is to ensure that condor common log does not get written to a shared filesystem. If the user knows for sure that the workflow submit directory is not on the shared filesystem, then they can opt to turn of the symlinking of condor common log file by setting the property pegasus.condor.logs.symlink to false.
Pegasus 3.0.0 Released
on November 29, 2010 with No Comments