WP1 WMS release 2 issues

1
WP1 WMS release 2 issues

• Massimo Sgaravatto
• INFN Padova

2
Outline

• Open bugs
• Missing functionalities
• Some other new things to do ?
• Problems in RPMs release process

3
Problems in RPMs release process

• It is happening again (as in release 1.x) that
  RPMs with big bugs (e.g. preventing a simple job
  submission) are released
• Different reasons
• Commit of important changes not properly tested
  just before producing RPMs
• Commit of changes requiring changes to other
  components just before producing RPMs

4
Problems in RPMs release process

• Different approach definitely needed
• E.g. something like
• No CVS commit allowed the day of the release
• half day of tests
• Better is via a test suite
• In this period allowed commits only to fix bugs
  found in these tests
• RPMs produced (via autobuild)

5
Known open issues

• Problem with edg-job-get-output which fails if
  not all Output SandBox files are available
• Problems with resubmission
• CEs already used are not considered
• Problem when proxy renewal register fails
• Unable to receive data error message
• Problem with sequence code in JobWrapper
• The sequence code for the JW is retrieved using
  edg_wl_GetSequenceCode() too early
• FileList problem
• Lock problems ?
• Integration with Optor (getaccesscost as rank)
• To be tested
• Problem with purger daemon
• Which proxy to use when querying the LB ??

6
Missing pieces

• Restart of daemons
• Should be done for all daemons (NS, WM, JC, LM,
  LB processes, proxy renewal daemon)
• GangMatching
• People are asking it
• Dynamic quota management in NS
• Interactive jobs
• Some modifications still needed in order to allow
  redirection of standard streams to pipes
• UI Man pages

7
Missing pieces

• Output Data Registration
• Registration of WMS services in RGMA and status
  scripts (?)
• BrokerInfo
• Software and documentation
• Documentation
• WMS user and administrator guide updated
• Build part missing (not urgent)
• JDL doc updated
• LB extended querying capability doc. missing
• Gangmatching note missing

8
Integration with VOMS

• Integration with VOMS
• VO not more in JDL/conf file but retrieved from
  proxy
• Modification in UI
• Matchamaking done wrt VO instead of UserSubject

9
Some other new things to do ?

• Exploit transfer_input_files, transfer_output_file
  s in Condor submit files to transfer sandbox
  files between WN and RB node via GASS instead of
  using gridftp ?
• Outbound IP connectivity wouldnt be needed
  anymore from WN to transfer sandboxes
• People keep asking to remove this constraint
• Exploit LB extended querying capabilities
• UI commands for these queries
• Possibility to define user tag in JDL to exploit
  extended querying capabilities
• Job wrapper in Fault Tolerant Shell (FTSH)
• To cope e.g. with transient failures (e.g.
  Globus-url-copy stuck or fails)
• E.g. Try 3 times to copy sandboxes - timeout 10
  minutes
• FTSH in VDT
• ...