SULU 58i

Summary

*  Improved error handling for failed replica lookups using DQ2 functions.

Friedrich Hönig discovered a case where this failure led to a complication

with using alternative file catalog hosts in case of missing replicas,

effectively preventing the fallback from working.

* Increased size of pilot TCP message buffer from 2kB to 4kB and

truncating too long messages sent from runJob subprocess to main pilot

which previously could cause a crash in the python TCP module (leading to

difficult to interpret payload failure). Problem discovered by Wolfgang

Ehrenfeld.

* Now accepting releases located in areas (/afs) that do not have

cmtconfig in their path. Requested by Wolfgang Ehrenfeld, to resolve an

issue with running jobs on CERN-RELEASE.

* Fixed a problem that prevented David Cameron from using ³/ruciotest² as

part of the file paths; pilot previously replaced /rucio-strings badly

from SURLs as part of the rucio path generation.

Version info

Removed Exception in getFileCatalogHosts() (ATLASExperiment)    [added to 59a]

Corrected undefined variable in bulkRegisterFiles() (SiteMover)    [added to 59a]

Truncating string in encode_string() to prevent potential problems with sending long TCP messages (pUtil)    [added to 59a]

Added error handling for empty replica dictionaries in getReplicaDictionaryRucio() (Mover)    [added to 59a]

Added special case for /cvmfs in getProperSiterootAndCmtconfig() (ATLASExperiment)    [added to 59a]

Cutting away the trailing /rucio bit from the destination path in getFullPath(), instead of a full replace of

/rucio in the path to prevent a problem with ruciotest paths used by David Cameron (SiteMover)    [added to 59a]

Increased buffer size from 2k to 4k in updateHandler::handle() (pilot)    [added to 59a]

Added warning message for too long TCP messages in updateJobInfo() (RunJobUtilities)    [added to 59a]

Now using full path including protocol for file catalog hosts in getFileCatalogHosts() (ATLASExperiment)    [added to 59a]

Removed exclusion of non-LFC hosts in getFileCatalogHosts() (ATLASExperiment)    [added to 59a]

Corrected host instead of thisExperiment.getFileCatalog() in call to getReplicaDictionaryRucio() in getReplicaDictionary() (Mover)    [added to 59a]

Added trackback info to exception handling in getRucioReplicaDictionary() (Mover)    [added to 59a]

Added trackback info to exception handling in getFileCatalogHosts() (ATLASExperiment)    [added to 59a]

Fixed left exception e string in getFileCatalogHosts() (ATLASExperiment)    [added to 59a]