Hello, How can we assign run_tags to new `AutoMate...
# ask-community
r
Hello, How can we assign run_tags to new
AutoMaterializePolicy
? In
1.2.X
, we used
build_asset_reconciliation_sensor
with
run_tags
to ensure reconciliation/automaterialize runs follows our run tags concurrency limits. I do not see how to do something similar with AutoMaterializePolicy.
1
Other question regarding reconciliation sensor and freshness policy: how can I enforce child asset with Freshness policy are rematerialized if any parent asset has changed ? It seems like reconciliation sensor only change status of child asset to "upstream data changed" without rematerialize it if a Freshness policy has been set on this child asset. The behavior I would like to have is to enforce both Freshness policy and rematerialization when any parent asset has changed.
s
@R Lucas this will be possible in our upcoming release this week: https://docs.dagster.io/master/concepts/assets/asset-auto-execution#run-tags (link is to unreleased docs)
1
Other question regarding reconciliation sensor and freshness policy: how can I enforce child asset with Freshness policy are rematerialized if any parent asset has changed ?
Have you had a chance to look at the docs? https://docs.dagster.io/concepts/assets/asset-auto-execution Basically, if you use an eager AutoMaterializePolicy, it will do it immediately, not based on the freshness policies.
r
this will be possible in our upcoming release this week: https://docs.dagster.io/master/concepts/assets/asset-auto-execution#run-tags (link is to unreleased docs)
@sandy from this documentation it seems like run tags can be set in
dagster.yaml
file which means that these tags would be global to all auto_materialize policies. Am I misunderstanding this or can auto_materialize runs be triggered by asset_reconciliation_sensor and associated run tags ?
s
from this documentation it seems like run tags can be set in
dagster.yaml
file which means that these tags would be global to all auto_materialize policies.
so your issue is that you want different concurrency limits for different assets?
r
Yes, Different groups of assets within our code locations are data extracted by their specific source. Concurrency is per data-source (i.e. API_1 concurrency limit is 5 while API_2 is 20). I managed this with reconciliation_sensor using run_tags parameter and defining a reconciliation sensor per data source (linked to one or multiple asset groups for this source).