Pegasus 4.7.3 Released

by on February 6, 2017 with No Comments

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.

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

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.

4,362 views