• Changed over from using Flux's network cache to using the Flux cluster table for scheduling work within a Flux cluster. This change, combined with some other changes made within this issue, improved Flux scheduling performance by over 25% and reduced the CPU load by 33%. This change requires a new column be added to the FLUX_CLUSTER table named "SCHEDULING". It is a BIT column. The upgrade scripts are provided below.
  • SQL Server: ALTER TABLE FLUX_CLUSTER ADD SCHEDULING BIT NULL;
  • MYSQL: ALTER TABLE FLUX_CLUSTER ADD SCHEDULING BIT;
  • Oracle: ALTER TABLE FLUX_CLUSTER ADD SCHEDULING NUMBER NULL;
• Resolved issues with CONCURRENCY_THROTTLE_CLUSTER_WIDE where there were instances when the throttle was not being adhered to.
• Resolved issue where once a workflow was executing, and it had multiple concurrent actions, its concurrency throttle could be exceeded.
• Improved the Flux scheduling logging (for debugging purposes) if the system property is set on the Flux JVM: "-Ddebug.chooser=true". This property is added in the setenv.sh/bat file if running Flux as a command line script, or in the engine.conf if starting Flux as a Windows service.

Correct issues with LDAP role assignments and login.

• Improved the LDAP authentication logging (for debugging purposes) if the system property is set on the Flux JVM: "-Dloginmodule.debug=true". This property is added in the setenv.sh/bat file if running Flux as a command line script, or in the engine.conf if starting Flux as a Windows service.
• Now, if a user's role is changed in LDAP, the next time the user signs in their new role will replace their existing role.
• Also - no longer allow users to login if there is no role defined for them. Before Flux would assign them as Workflow Masters.

Added expandable/collapsible audit trail table to Dashboard.

Corrected issue where the CLUSTER_WIDE_CONCURRENCY_THROTTLE setting was not being honored. If this setting is required, make sure the following engine configuration settings are present

• CLUSTER_NETWORKING_ENABLED=FALSE
• CACHE_TYPE=NONE

Mitigate SQL Deadlocks in MySql during attemptSeal causing delays in processing throughput. Note that this requires the creation of a two new database indexes - see /FLUX_HOME/doc/upgrade-8.0.13-mysql.sql. Need to add the following 2 indexes:

CREATE INDEX FLUX_READY8 ON FLUX_READY(FLOW_CHART_NAME,STATE);
CREATE INDEX FLUX_READY9 ON FLUX_READY(FLOW_CHART_NAME,STATE,ACTION);

Allow namespacing the assignment of agent pools to ensure users only can process their actions on allowed pools. Allow defining runtime variables for each agent pool to show in the agent pool dialogs in the Cockpit user interface.

• If no runtime variables are defined for agent pools, display the currently connected agents in an action's agent pool dropdown dialog.
• If runtime variables are defined for agent pools, display the defined agent pool runtime variables in an action's agent pool dropdown dialog. If agent pool runtime variables are defined any currently connected agents will not be displayed in an action's agent pool dropdown dialog.
• In either of the above cases, If an agent pool is already defined in an existing workflow but is not currently available, display it in an action's agent pool dropdown dialog too so prior agent pool assignments in already defined workflows are not lost.

• Sample agent pool runtime variables would be:
  /AGENTPOOL_1=Denied
  /AGENTPOOL_2=Denied
  /Developer/AGENTPOOL_1=DeveloperPool1
  /Prod/AGENTPOOL_1=ProdPool1

  In the agent pool dropdown dialogs only the text AGENTPOOL1 and AGENTPOOL2 would display. The actual agent pools that the actions execute within would be resolved at runtime, so a job submitted into the /Developer namespace would use the runtime variable value resolved to by /Developer/AGENTPOOL_1 (which is DeveloperPool1), and jobs submitted to /Prod would use the runtime variable value resolved to by /Prod/AGENTPOOL_1 (which is ProdPool1).

Corrected issue where adding a business calendar in a prescript - with LDAP enabled - was failing in a security check.

Added a business calendar to support US Federal Reserve Holidays - which differ in processing on Veterans Day with the US Federal Holidays calendar.

If a file action fails to connect to a remote server, or fails to transfer a file due to a network error - retry the connect and the transfer a configurable number of times, with a configurable delay between each retry. The default is no retries with a 30 second delay. This can be configured in the runtime configuration properties files at the root level (i.e., for all remote file servers) or for specific remote file servers using their IP address or hostname. For example:

/FILERETRYDELAY=45
/FILERETRYCOUNT=3
/CONNECTRETRYDELAY=60
/CONNECTRETRYCOUNT=3
/DOC.FLUX.LY/FILERETRYDELAY=45
/DOC.FLUX.LY/FILERETRYCOUNT=3
/DOC.FLUX.LY/CONNECTRETRYDELAY=60
/DOC.FLUX.LY/CONNECTRETRYCOUNT=3

Fix issue where Cockpit was only rendering the first form (on a manual trigger) in a workflow.

Cockpit Fix for audit trail tab throwing invalid parameter on first access from Dashboard

The trigger now exits with an empty result set if no file is present and the minimum count is 0.

BUG

Correct signout and login issue related to Firefox - getting a "Could not fetch PGP keys info" and "Could not fetch calendar info" errors on signout

Corrected issue where agent names instead of agent pool names were displayed in action dialogs

Stopped some instances where the Cockpit Status Viewer was attempting to show audit trail entries if an action was double clicked in the Status Viewer