Add a new worker/queue to avoid waiting for mask recalculation. And fix celery...

Add a new worker/queue to avoid waiting for mask recalculation. And fix celery acks late setting typo.

app/escriptorium/settings.py

@@ -181,19 +181,21 @@ CELERY_RESULT_BACKEND = 'redis://%s:%d' % (REDIS_HOST, REDIS_PORT)
 CELERY_ACCEPT_CONTENT = ['application/json']
 CELERY_TASK_SERIALIZER = 'json'
 CELERY_RESULT_SERIALIZER = 'json'
-CELERYD_ACKS_LATE = True
+CELERY_ACKS_LATE = True
 
 # time in seconds a user has to wait after a task is started before being able to recover
 TASK_RECOVER_DELAY = 60 * 60 * 24  # 1 day
 
 CELERY_TASK_QUEUES = (
     Queue('default', routing_key='default'),
+    Queue('live', routing_key='live'),  # for everything that needs to be done on the spot to update the ui
     Queue('low-priority', routing_key='low-priority'),
-    Queue('gpu', routing_key='gpu'),
+    Queue('gpu', routing_key='gpu'),  # for everything that could use a GPU
 )
 CELERY_TASK_DEFAULT_QUEUE = 'default'
 CELERY_TASK_ROUTES = {
     # 'core.tasks.*': {'queue': 'default'},
+    'core.tasks.recalculate_masks': {'queue': 'live'},
     'core.tasks.generate_part_thumbnails': {'queue': 'low-priority'},
     'core.tasks.train': {'queue': 'gpu'},
     'core.tasks.segtrain': {'queue': 'gpu'},

docker-compose.yml

@@ -73,6 +73,10 @@ services:
       - OMP_NUM_THREADS=1
     command: "celery worker -l INFO -E -A escriptorium -Ofair --prefetch-multiplier 1 -Q default -c ${CELERY_MAIN_CONC:-10} --max-tasks-per-child=10"
 
+  celery-live:
+    <<: *app
+    command: "celery worker -l INFO -E -A escriptorium -Ofair --prefetch-multiplier 1 -Q live -c ${CELERY_LIVE_CONC:-10} --max-tasks-per-child=10"
+
   celery-low-priority:
     <<: *app
     command: "celery worker -l INFO -E -A escriptorium -Ofair --prefetch-multiplier 1 -Q low-priority -c ${CELERY_LOW_CONC:-10} --max-tasks-per-child=10"