5. Pegasus 4.7.x Series

5.1. Pegasus 4.7.5

Release Date: September 5, 2017

We are happy to announce the release of Pegasus 4.7.5 . Pegasus 4.7.5 is a minor release, which contains minor enhancements and fixes bugs. This will most likely be the last release in the 4.7 series, and unless you have specific reasons to stay with the 4.7.x series, we recommend to upgrade to 4.8.0.

The following issues were addressed and more information can be found in the Pegasus Jira (https://jira.isi.edu/)

5.1.1. Improvements

  1. [PM-1191] - If available, use GFAL over globus url copy #1305

    JGlobus is no longer actively supported and is not in compliance with RFC 2818 (https://docs.globus.org/security-bulletins/2015-12-strict-mode). As a result cleanup jobs using pegasus-gridftp client would fail against the servers supporting the strict mode. We have removed the pegasus-gridftp client and now use gfal clients as globus-url-copy does not support removes. If gfal is not available, globus-url-copy is used for cleanup by writing out zero bytes files instead of removing them.

  2. [PM-1146] - There doesn’t seem to be a way to get a persistent URL #1260 for a workflow in dashboard

  3. [PM-1186] - pegasus-db-admin should list compatibility with latest #1300 pegasus version if no changes to schema

  4. [PM-1187] - make scheduler type case insensitive for grid gateway #1301 in site catalog

5.1.2. Bugs Fixed

  1. [PM-1032] - Handle proper error message for non-standard python #1146 usage

  2. [PM-1171] - Monitord regularly produces empty stderr and stdout #1285 files

  3. [PM-1172] - pegasus-rc-client deletes all entries for a lfn #1286

  4. [PM-1173] - cleanup jobs failing against Titan gridftp server due #1287 to RFC 2818 compliance

  5. [PM-1176] - the job notifications on failure and success should #1290 have exitcode from kickstart file

  6. [PM-1181] - monitord fails to exit if database is locked #1295

  7. [PM-1185] - destination in remote file transfers for inter site #1299 jobs point’s to directory

  8. [PM-1193] - “pegasus-rc-client list” modifies rc.txt #1307

  9. [PM-1196] - pegasus-statistics is not generating jobs.txt for some #1310 large workflows

  10. [PM-1207] - Investigate error message: Normalizing ‘4.8.0dev’ to #1321 ‘4.8.0.dev0’

  11. [PM-1208] - Improve database is locked error message #1322

  12. [PM-1209] - Analyzer gets confused about retry number in #1323 hierarchical workflows

  13. [PM-1211] - DAX API should tell which lfn was a dup #1325

  14. [PM-1213] - pegasus creates duplicate source URL’s for staged #1327 executables

  15. [PM-1217] - monitord exits prematurely, when in dagman recovery mode #1331

5.2. Pegasus 4.7.4

Release Date: February 27, 2017

We are pleased to announce Pegasus 4.7.4. Pegasus 4.7.4 is a minor release of Pegasus and includes improvements and bug fixes to the 4.7.4 release.

5.2.1. Bugs Fixed

  1. [PM-1148] - kickstart should print a more helpful error message if #1262 the executable is missing

  2. [PM-1160] - Dashboard is not recording the hostname correctly #1274

  3. [PM-1163] - Confusing error message in pegasus-kickstart #1277

  4. [PM-1164] - worker package in submit directory gets deleted during #1278 workflow run

5.3. Pegasus 4.7.3

Release Date: February 6, 2017

We are happy to announce the release of Pegasus 4.7.3. Pegasus 4.7.3 is a minor release of Pegasus and includes improvements and bug fixes to the 4.7.2 release. It has a bug fix without which monitoring will break for users running with HTCondor 8.5.8 or higher.

5.3.1. Improvements

  1. [PM-1109] – dashboard to display errors if a job is killed instead #1223 of exiting with non zero exitcode

    pegasus-monitord did not pass signal information from the kickstart records to the monitoring database. If a job fails and because of a signal, it will now create an error message indicating the signal information, and populate it.

  2. [PM-1129] – dashboard should display database and pegasus version #1243

  3. [PM-1138] – Pegasus dashboard pie charts should distinguish between #1252 running and unsubmitted

  4. [PM-1155] – remote cleanup jobs should have file url’s if possible #1269

5.3.2. Bugs Fixed

  1. [PM-1132] – Hashed staging mapper doen’t work correctly with sub #1246 dax generation jobs

    For large workflows with dax generation jobs, the planning broke for sub workflows if the dax was generated in a hashed directory structure. It is now fixed.

    Note: As a result of this fix, pegasus-plan prescripts for sub workflows in all cases, are now invoked by pegasus-lite

  2. [PM-1135] – pegasus.transfer.bypass.input.staging breaks symlinking #1249 on the local site

  3. [PM-1136] – With bypass input staging some URLs are ending up in #1250 the wrong site

  4. [PM-1147] – pegasus-transfer should check that files exist before #1261 trying to transfer them

    In case where the source file url’s don’t exist, pegasus-transfer used to still attempt multiple retries resulting in hard to read error messages. This was fixed, whereby pegasus-transfer does not attempt retries on a source if a source file does not exist.

  5. [PM-1151] – pegasus-monitord fails to populate stampede DB #1265 correctly when workflow is run on HTCondor 8.5.8

  6. [PM-1152] – pegasus-analyzer not showing stdout and stderr of #1266 failed transfer jobs

    In case of larger stdout+stderr outputted by an application, we store only first 64K in the monitoring database combined for a single or clustered job. There was a bug whereby if a single task outputted more than 64K nothing was populated. This is fixed

  7. [PM-1153] – Pegasus creates extraneous spaces when replacing <file #1267 name=”something” />

    DAX parser was updated to not add extraneous spaces when constructing the argument string for jobs

  8. [PM-1154] – regex too narrow for GO names with dashes #1268

  9. [PM-1157] – monitord replay should work on submit directories that #1271 are moved

    pegasus generated submit files have absolute paths. However, for debugging purposes where a submit directory might be moved to a different host, where the paths don’t exist. monitord now searches for files based on relative paths from the top level submit directory. This enables users to repopulate their workflow databases easily.

5.4. Pegasus 4.7.2

Release Date: November 22, 2016

We are happy to announce the release of Pegasus 4.7.2. Pegasus 4.7.2 is a minor release of Pegasus and includes improvements and bug fixes to the 4.7.1 release.

5.4.1. Improvements in 4.7.2 are

  • [PM-1141] - The commit to allow symlinks in pegasus-transfer broke #1255 PFN fall through

  • [PM-1142] - Do not set LD_LIBRARY_PATH in job env #1256

  • [PM-1143] - R DAX API #1257

  • [PM-1144] - pegasus lite prints the wrong hostname for non-glidein #1258 jobs

5.5. Pegasus 4.7.1

Release Date: November 7, 2016

We are happy to announce the release of Pegasus 4.7.1. Pegasus 4.7.1 is a minor release of Pegasus and includes improvements and bug fixes to the 4.7.0 release.

5.5.1. Improvements in 4.7.1 are

  • Fix for stage in jobs with repeated portion of LFN [PM-1131] #1245

  • Fix for pegasus.transfer.bypass.input.staging breaks symlinking on the local site [PM-1135] #1249

  • Capture the execution site information in pegasus lite [PM-1134] #1248 ]

  • Added ability to check CVMFS for worker package

5.6. Pegasus 4.7.0

Release Date: October 14, 2016

We are happy to announce the release of Pegasus 4.7.0. Pegasus 4.7.0 is a major release of Pegasus and includes all the bug fixes and improvements in the 4.6.2 release

New features and Improvements in 4.7.0 are

- automatic submit directory organization
- improved directory management on staging site in nonsharedfs mode
- R DAX API
- pegasus-analyzer reports information about held jobs
- check for cyclic dependencies in DAG

5.6.1. New Features

  1. [PM-833] – Pegasus should organize submit files of workflows in #951 hierarchal data structure

    Pegasus now automatically distributes the files in HTCondor submit directory for all workflows in 2 level directory structure. This is done to prevent having too many workflow and condor submit files in one directory for a large workflow. The behavior of submit directory organization can be controlled by the following properties

    pegasus.dir.submit.mapper.hashed.levels the number of directory levels used to accomodate the files. Defaults to 2. pegasus.dir.submit.mapper.hashed.multiplier the number of files associated with a job in the submit directory. defaults to 5.

    Note that this is enabled by default. If you want to have pre 4.7.0 behavior you can pegasus.dir.submit.mapper Flat

    Submit mapper properties are documented in the user guide here https://pegasus.isi.edu/docs/4.7.0/properties.php#site_dir_props

  2. [PM-833] – Pegasus should manage directory structure on the staging #951 site

    For non sharedfs mode, Pegasus will now automatically manage the directory structure on the staging site in a hierarchal directory structure via use of staging mappers. The staging mappers determine what sub directory on the staging site a job will be associated with. Before, the introduction of staging mappers, all files associated with the jobs scheduled for a particular site landed in the same directory on the staging site. As a result, for large workflows this could degrade filesystem performance on the staging servers. More information can be found in the documentation at https://pegasus.isi.edu/docs/4.7.0/ref_staging_mapper.php

  3. [PM-1036] – R DAX API #1150

    Pegasus now includes an R API for generating DAXes of complex and large workflows in R environments. The API follows the Google’ R style guide, and all objects and methods are defined using the S3 OOP system. The source package can be obtained by running pegasus-config –r or from the Pegasus’ downloads page. A tutorial workflow can be generated using pegasus-init, and an example workflow is provided in the examples folder. More information can be found in the documentation at https://pegasus.isi.edu/docs/4.7.0/dax_generator_api.php#api-r

    Related JIRA item: [PM-1074] – Add R example to pegasus-init #1188

  4. [PM-1126] – pegasus-analyzer should report information about held jobs #1240

    Pegasus monitoring daemon now populates the reason for held jobs in it’s database. Both pegasus-analyzer and dashboard were updated to show this information. Related JIRA items: [PM-1121] – Store reasons for workflow failure in stampede database #1235 [PM-1122] – update dashboard to display reasons for workflow state and jobstate #1236 [PM-1058] – Create homebrew tap for pegasus #1172

  5. Pegasus is now also available via home-brew on MACOSX via a tap repository github.com/pegasus-isi/homebrew-tools

    It contains formulas for pegasus and htcondor.Users can do:$ brew tap pegasus-isi/tools

$ brew install pegasus htcondor $ brew tap homebrew/services $ brew services start htcondor

  1. [PM-928] – pegasus-exitcode should write its output to a log file #1045

    pegasus-exticode is now set to write to a workflow global log file ending in exitcode.log that captures pegasus-exitcode stdout and stderr as json messages. This allows users to check pegasus-exitcode messages, which otherwise would have been set to /dev/null by condor dagman.

  2. [PM-1115] – Pegasus to check for cyclic dependencies in the DAG #1229 Pegasus now explicitly checks for cyclic dependencies and reports one of the edges making up the cycle.

  3. [PM-1054] – Add option to ignore files in libinterpose #1168

    kickstart now has support for environment variables KICKSTART_TRACE_MATCH and KICKSTART_TRACE_IGNORE that determine what file accesses are captured via lib interpose. The MATCH version only traces files that match the patterns, and the IGNORE version does NOT trace files that match the patterns. Only one of the two can be specified.

  4. [PM-915] – modify kickstart to collect and aggregate runtime #1033 metadata

  5. [PM-1004] – update metrics server ui to display extra planner configuration metrics #1121

  6. [PM-1111] – pegasus planner and api’s should have support for #1225 ppc64 as architecture type

  7. [PM-1117] – Support for tutorial via pegasus-init on bluewaters #1231

5.6.2. Improvement

  1. [PM-1125] – Disable builds for older platforms #1239

  2. [PM-1116] – pass task resource requirements as environment #1230 variables for job wrappers to pick up

  3. [PM-1112] – enable variable expansion for regex based replica #1226 catalog

  4. [PM-1105] – Mirror job priorities to DAGMan node priorities #1219

    HTCondor ticket 5749 . We can assign DAG priorities only if detected condor version is greater than 8.5.

  5. [PM-1094] – pegasus-dashboard file browser should load on demand #1208

  6. [PM-1079] – pegasus-statistics should be able to skip over failures #1193 when generating particular type of stats

  7. [PM-1073] – condor_q changes in 8.5.x will affect pegasus-status #1187

  8. [PM-1023] – KIckstart stdout/stderr as CDATA #1138

  9. [PM-749] – Store job held reasons in stampede database #867

  10. [PM-1088] – Move to relative paths in dagman and condor submit #1202 files

  11. [PM-900] – site catalog for XSEDE #1018

  12. [PM-901] – site for OSG #1019

5.6.3. Bugs Fixed

  1. [PM-1061] – pegasus-analyzer should detect and report on failed job #1175 submissions

  2. [PM-1118] – database changes to jobstate and workflow state tables #1232

  3. [PM-1124] – Hashed Output Mappers throw unable to instantiate error #1238

  4. [PM-1127] – Wf adds worker package staging even though it has already placed a worker package in place #1241