We are delighted to announce the release of Cylc 8.6.0
This release contains one particularly notable change, Cylc IDs are no longer limited to matching n=0 tasks. For example, this means that cylc hold //1234/FAMILY will now hold all members of FAMILY not just the members that are presently held in the scheduler’s window.
The minimum Python version for all Cylc and Rose components has increased to 3.12.
It marks the point where Cylc 8 definitively eclipses any advantage that Cylc 7had.
(That’s specifically with regard to runtime interventions - Cylc 8 was better from the outset in almost every other respect).
For nostalgic Cylc 7 users: by sheer luck the inefficient (by Cylc 8 standards) Cylc 7 scheduler often made runtime interventions easier than might be expected, because it kept many more tasks in memory. Going beyond this “active window” of the workflow could be very difficult in Cylc 7, but often wasn’t necessary.
With task matching beyond the n=0active window in Cylc 8.6 we can now do all the powerful Cylc 8 interventions without worrying about whether the tasks involved are still in memory or not.
Relatedly, we are gradually adding specific advice to the User Guide Interventions section.
There are no changes that need to be made to workflow configurations - so just install Cylc 8.6 along side 8.5, and you’re good to go.
As advised here Installation — Cylc 8.6.0 documentation we recommend that you install a wrapper called cylc (just use the provided one and modify it a bit) that is in the default $PATH and selects which version to run based on environment (e.g., the latest version by default, or $CYLC_VERSION). Each separate version should be installed as a complete conda environment, with everything needed in it.
(Let us know if the instructions on that are not clear enough).
Thanks @hilary.j.oliver ! Yes, we are using the wrapper cylc script, so it sounds like it will be smooth. Do we need to stop and start workflows? Or do they carry over seamlessly?
With the wrapper script approach, you can add new Cylc deployments without affecting running workflows. Restart your workflows to pick up the latest version:
cylc stop --now --now <workflow-id>
cylc play <workflow-id>
We typically make the latest deployment the default one. Otherwise, use the CYLC_VERSION environment variable to change which version you are using. Cylc may prompt you to confirm the version upgrade, use the --upgrade option to bypass this.
We have released cylc-flow 8.6.1, cylc-uiserver 1.8.1
These releases contain a number of bug fixes as well as some enhancements to the GUI, highlights:
The tutorial workflows have changed data source (the old tutorials will become defunct shortly).
The workflows sidebar in the Cylc GUI has been redesigned. The task state totals now reflect only the n=0 window As a result, they don’t change when the window is adjusted, and handled task failures are omitted from the totals.
The memory usage of both cylc-flow and cylc-uiserver should have improved in certain situations.
Fixed issues related to cylc trigger and platform selection in paused workflows.
Python 3.14 support.
A metomi-rose bugfix release will follow shortly.
How to update installations
How to update site installations
Sites should create a new environment for this release and manage parallel installations using the provided wrapper script .
How to update stand-alone installations
# First shut down running workflows:
cylc stop '*'
# For Conda managed environments:
conda install 'cylc-flow=8.6.1' 'cylc-uiserver=1.8.1'
# For Pip managed environments:
pip install 'cylc-flow==8.6.1' 'cylc-uiserver==1.8.1'
We have released cylc-flow 8.6.3, cylc-uiserver 1.8.4
Notable bugfixes:
A memory leak in cylc-flow has been addressed which could cause the scheduler of long-lived cycling workflows with branched graphs to slowly accumulate memory.
The top-level retry indicators in the GUI will now show for tasks which belong to families.
Fixed a bug where tasks would sometimes disappear from the table view.
We have released metomi-rose 2.7.0 and cylc-rose 1.7.1
The Rose release has two significant changes:
The rose edit GUI has been restored after being ported from Python 2 / GTK 2.
Conda-forge installation:
The rose edit GUI is now part of the default metomi-rose Conda package.
For installations where rose edit is not required, use the metomi-rose-base package.
PyPi installation:
The rose edit GUI is not included in the default package, install Rose with the “edit” optional extra e.g, pip install rose[edit].
It you choose to install via PyPi you will need to install system dependencies manually, e.g, GTK3+ and Cairo. More info in the Rose installation documentation.
The rosie disco server has also been restore after porting.
Conda-forge installation:
This is not available via Conda at present, let us know if this would be useful.
PyPi installation:
rosie disco is not included in the default package.
Install Rose with the “disco” optional extra, e.g, pip install rose[disco].
The newly ported rose edit should be considered an experimental feature while teething issues are ironed out. Please report any issues.
This release includes some small updates to the GUI:
Task “modifier icons” (e.g, held, retry, xtriggered, etc) now show on family icons as well as task icons.
Some small bugfixes.
But this release also includes Cylc Review. Users of Cylc 7 may remember Cylc Review as a public web app for viewing workflows:
It can display the status of tasks which are no longer visible in the GUI.
Provides shareable URLs for job log files and more.
It is suitable for use in trusted environments where users are permitted to view other user’s workflows.
Cylc Review has now been ported to Cylc 8 where it runs as Jupyterhub Service. It is not enabled by default, but can be activated by administrators by adding a few lines of configuration, details in the README:
