Home
Diffusion Flamenco Manager (hotfix-v2.3) Browse

flamenco-manager/CHANGELOG.md
ce28b0a78182hotfix-v2.3

Last Change
Raw File

CHANGELOG.md
View Options

Changelog for Flamenco Manager
==============================
## Version 2.0.16 (in development)
- Always log which version of Flamenco is running.
## Version 2.0.15 (released 2017-09-09)
- Flamenco Manager can now be run from a different directory than the executable is in. It searches
for web templates both relative to the current working directory and relative to the executable's
own directory.
## Version 2.0.14 (released 2017-09-07)
- Fixed panic when enabling UPnP/SSDP auto-discovery on Windows.
- Bundled MongoDB itself with Flamenco Manager, so that a separate MongoDB installation is no longer
required. When the `database_url` in `flamenco-manager.yaml` is empty, this bundled database
server will be used. Note that running unit tests while developing Flamenco Manager still requires
a separate server instance.
- Added web interface for configuring Flamenco Manager. Start it using `flamenco-manager -setup`.
- Restyled the dashboard.
## Version 2.0.13 (released 2017-07-04)
- Added auto-discovery via UPnP/SSDP, so that Workers can automatically find this Manager on the
network. This can be turned off by setting `ssdp_discovery` to `false`. For now only discovery via
IPv4 is supported; this deficiency is reported at https://github.com/fromkeith/gossdp/issues/4,
and also for the alternative UPnP/SSDP implementation at https://github.com/koron/go-ssdp/issues/4
## Version 2.0.12 (released 2017-06-23)
- Added `-purgequeue` CLI command, which erases all queued task updates from the local MongoDB, then
exits Flamenco Manager.
- The `-cleanslate` CLI command now exits immediately when there are no tasks locally cached, i.e.
when it would be a no-op anyway.
- The `-purgequeue` and `-cleanslate` commands now show how many items they would erase, before
asking the user for confirmation.
## Version 2.0.11 (released 2017-06-16)
- Fixed a compatibility issue with Windows 10.
## Version 2.0.10 (released 2017-06-09)
- Added --factory-startup option to example Blender CLI variable. This is needed because this
option was removed from the hard-coded values in the Flamenco Worker.
- When worker asks for tasks, also check already assigned tasks.
[T51519](https://developer.blender.org/T51519)
- Changes the way durations are stored in flamenco_manager.yaml. They must now be expressed
including units with a suffix (h, m, or s), rather than having the units in the configuration
variable name.
OLD: download_task_sleep_seconds = 30
NEW: download_task_sleep = 30s
- Introduced path replacement variables, to allow Clients and Workers to run on different platforms.
## Version 2.0.9 (released 2017-05-11)
- Workers: only store host part of worker's address, and not the port number.
- Dashboard: Shorten task and worker IDs and nicer timestamp formatting.
## Version 2.0.8 (released 2017-05-09)
- Use mutex in scheduler to avoid race condition.
- Clear the worker's current task upon sign-on. This makes the dasboard less confusing
when the worker's task was rescheduled to another worker.
- On the dashboard, for the current/task of a worker, show the last timestamp/status
that this particular worker worked on it (rather than showing the last timestamp/status
of the task itself)
- Dashboard: Show workers as table (instead of blocks).
- Dashboard: click on worker ID or address to copy it to the clipboard.
## Version 2.0.7 (released 2017-04-26)
- Added 'Kick task downloader' button to dashboard.
- Dashboard: also update Manager version from JSON.
## Version 2.0.6 (released 2017-04-21)
- Reduced logging of workers requesting tasks, as we now have a nice dashboard
to show the status.
- Reduced logging of task downloading.
- Not-seen workers are now moved to "Old workers" after 14 days (instead of 31)
## Version 2.0.5 (released 2017-04-20)
- Fixed race condition where a cancelled task could be re-activated
by a worker.
## Version 2.0.4 (released 2017-04-07)
- Small dashboard JS tweak: hide workers we haven't seen in over a
month.
- Added support for task types. Only tasks of a type that workers
support will be scheduled for them. This also adds a /sign-on
URL so workers can send a current list of supported task types
(and their current nickname) to the Manager.
- Dashboard: Vertically align last-rendered image.
## Version 2.0.3 (released 2017-04-04)
- Implemented error checking for JSON encoding & sending.
This should work around a timeout issue we've seen, where a worker
times out while waiting for the scheduler. The Manager would ignore
the error and keep the task assigned to the worker. Now it detects
the error and unassigns the worker.
## Version 2.0.2 (released 2017-03-31)
- Added `/output-produced` endpoint, where Workers can register the files
they have produced (e.g. rendered images). This allows the "last rendered
image" feature to work on filesystems that don't support inotify as well.
## Version 2.0.1 (released 2017-03-30)
- Added display of last-rendered image in a certain directory. Use
the `watch_for_latest_image` configuration file option to enable this.
## Version 2.0
- Initial version for this changelog.