Changes in 2.2 1714849
- RenderMan (prman) progress messages are now detected correctly by tractor-blade when Katana (renderboot) wraps them in additional logging text.
- Jobs submitted to tractor-engine from clients using the Python job authoring API method EngineClient.spool() now correctly abide by the current tractor.config setting AllowJobOwnerOverride. Note that this fix requires clients to re-import the new tractor.api.author module.
- Address a tractor.engine threading issue that could cause large “config” backlogs and slow dispatching in the unusual case where tq scripting clients make requests to the engine as usual, but site network routing issues prevented engine reply buffers from being delivered back to a few of those clients. Now those stalled deliveries time-out without affecting other transactions.
- Additional internal handling of the "assigner Cmd not Ready” state inconsistency condition that can arise in some cases of simultaneous task retry and job restart.
- Fix negative elapsed times that were sometimes displayed in the Dashboard as tasks finished, prior to a display refresh.
Changes in 2.2 1677499
- Custom menu items will cause a new window to be opened if there is any script output, even if the menu item is configured to normally suppress a new window. This enables problematic scripts to be more easily detected and debugged.
- Custom menu items support "login" as a special entry in the "values" list, which will cause the Dashboard user name to be a part of the payload sent to the menu item's script.
- Custom menu items will now observe user names and "@owner" in the crews attribute forjoband task custom menu items.
- Fixed bug in which selecting next task bystatein Dashboard task list was not automatically scrolling totask.
- Fixed the calculation of the elapsed time of a skipped task in the rollover of the Dashboard job graph.
- Sorting of tasks in the Dashboard task list has been corrected.
- Dashboard preview commands can now be displayed and run for archived jobs.
- The kill operation is now supported intqand the query API. It is used to kill a running command and leaves the task in an error state.
- tqhas a newreloadconfigcommand to enable the triggering of configuration file reloading from the command line.
- The query API will only return non-registered blades when the archive flag is set to True.
- tractor-dbctl--exec-sqlnow emits error messages.
- The blade caches user database entries in order to be more resilient against transient LDAP server outages.
- The systemd configuration directory defaults to the Tractor installation config/ directory, consistent with the sysvinit setting.
- systemd now starts tractor blade as root by default, consistent with the sysvinit setting.