GitLab

Improve SPEL expressions

 - ability to modify variable models
 - can use temp and tempMap variables to store and reuse some data in the expression
 - can validate or invalidate an expression with the 'valid' variable
 - utility functions which facilitate sequencing of instructions

Add configurable http request header size

Change package.loaded location

Task fork : Add retry mechanism to create context file

synchronize fork

Additionnally, increase the default start task timeout

Add configurable walltime

 - global walltime configuration at scheduler level
 - generic information configuration (can be applied to workflows or tasks)
 - walltime configuration priority is task-defined > generic-info-defined > globally-defined

upgrade docker compose engine version

Start scheduler in frozen|paused|stopped state

handle empty passwords

as we observed a case where an empty password could wrongly allow arbitrary connections from user to some AD server, we need to prevent empty passwords in the ldap module code.

upgrade docker compose engine version

upgrade docker compose engine version

Add rest urls as variables for most services

 - system variables for cloud-automation, job-planner and notification-service
 - configurable public version of these urls (for example if behind a reverse proxy or with different domain)

Improve logging in Scheduler Frontend.

Add log messages regarding nature of requests received and originator.

For example, it is important to know who asks to kill a job, stop the scheduler, etc.

Increase TaskCreationTimerTest maximum time

Improve SchedulerFrontendState concurrency

This improvement follows a dead lock observed in SchedulerFrontendState
 - use ConcurrentHashMap instead of global locks for user sessions. This reduces the lock complexity by having only a global schedulerStateLock
 - Additinnally, improve token checks in SelectionManager (due to a bug observed at Creos)
 - Add missing null checks on jobState
 - Use ReentrantReadWriteLock instead of synchronized for ClientJobState

Upgrade HSQLDB version

As hsqldb now behaves like other databases, we must now provide maximum length for all lob columns.

Improve Deployment Index handling

Previous implementation was using an index counter always increasing.

Problem: this generate log files and data files which are big and can consume the disk space over time.

Solution: use a set of deployed indexes instead of a counter and find the first available index absent from this set

Fill updated Variables and Models even when the workflow variables are not valid