When a flow start, his flow steps have lower priority than already scheduled jobs, hence steps are only executed when the queue is drained which considerably increase flow execution time when queue size is significant.
This add a very low priority to flow steps in order to prioritize finishing a started flow over new jobs.
* feat(cache): remove persistent raw values from queue
* test: add tests and fix discovered regressions
* fix csharp executor after rebase
* invalidate script cache after locking
* fix lite flow version race
* fix flow dependencies lock race
* update version flag to `v.1.440`
* feat(backend): implement flow scripts
* conditionally resolve reduced flow value
* conditionally use flow scripts based on the min worker version
* grant all to `flow_{version_lite,node}`
* do not silently discard errors
* env: add flow script feature flag
* flow script on cloud automatically
---------
Co-authored-by: Ruben Fiszel <ruben@rubenfiszel.com>
* init
* croner with fall back on scheduling failure
* adding packages
* fallback mechanism v1
* removing debug console log
* improving cargo organization
* queue cargo.toml
* use workspace cargo-tz
* test next occurrence when creating new croner schedule
* toggle label
* treat legacy edit requests where no cron_version is provided as v1
* fix edit schedule
* Allow setting progress explicitly from script body.
This feature exposes:
* `getProgress`
* `setProgress`
* `incProgress`
API in TypeScript client (python is coming soon).
NOTE: Progress cannot be out of range 0..100 and cannot decrease.
With exposed APIs there is also UI changes, so progress can be shown for individual jobs as well.
For optimization reasons, jobs start to ask for progress only after N-seconds of execution.
* feat: Add `shell.nix`
If you dont have anything but nix, dont worry, run nix-shell in root, or activate with direnv and get all needed dependencies
NOTE: You will still need docker
* feat: Add `dev.nu` to typescript client
Little helper function, allowing developer to work on ts client easier.
To use:
`./dev.nu watch`
Now add import of windmill in body of your script and `//nobundle` on top of the file
Edit ts client in your favourite editor and hit save. Script will do the rest.
* Cleanup files
* Fix: Failed to deserialize query string: missing field `get_progress`
* perf: Implement non-naive polling mechanism for getting job progress
* Add independant delay for getProgress
Problem in `TestJobLoader`:
There should be 2 delays:
One until we find our first progress (every 5s)
Once we found our first progress, we can do it every second
* nit: Use `query_scalar!` instead of `query_as`
* Fix: Sql error, no rows returned by a query that expected to return at least one row
* refactor: Remove global CSS for JobProgressBar
* Change UI for progress of flow subjobs
* Replace `Step 1` with `Running` in ProgressBar for individual jobs
* Remove `incProgress`
incProgress is not very usefull and error-prone
* perf: Set metric only for jobs that are actually using it
(https://github.com/windmill-labs/windmill/pull/4373#discussion_r1759843773)
* Offload registering progress from clients to server
* Add `jobId?` argument to typescript-client's `setProgress` and `getProgress`
Allows to set progress of other jobs and flows,
if jobId specified, than flow id will be inferred automatically.
Could be used by SDK.
* Add `Error::MetricNotFound` for better error handling
* Fix: Make `JobProgressBar` display in red when failed
* Add persistant progress bar
Now you can reload the page after job is done and progress will be still there
* Allow succeeded individual job's progress bar stick to 100%
* Add python support
* nit: Remove usage of undefined variable in python-client
* Add `async` in ts client (for error handling)
* nit(frontend): Remove unused import
* Dont load JobProgressBar when it is not needed
* nit: npm check fix
* cargo sqlx prepare
* fix sqlx
---------
Co-authored-by: Ruben Fiszel <ruben@rubenfiszel.com>
