68.1

DO NOT USE! Due to a security bug in an imported library this version of Cromwell should NOT be used. Its release files have been deleted.

68

68 Release Notes

DO NOT USE! Due to a a security bug in an imported library this version of Cromwell should NOT be used. Its release files have been deleted.

Virtual Private Cloud

Previous Cromwell versions allowed PAPIV2 jobs to run on a specific subnetwork inside a private network by adding the
information to Google Cloud project labels.

Cromwell now allows PAPIV2 jobs to run on a specific subnetwork inside a private network by adding the network and
subnetwork name directly inside the virtual-private-cloud backend configuration. More info
here.

67

67 Release Notes

Configuration updates for improved scaling

Some configuration changes were introduced in Cromwell 67 to support improved scaling. See Cromwell's reference.conf for details on new parameters.

• I/O throttling moved from io to its own io.throttle stanza; config updates may be required if these values are currently being overridden in local deployments.

• The default system.job-rate-control has been changed from 50 per second to 20 per 10 seconds.

• New configuration parameters have been introduced for values which were previously hardcoded constants:

  • system.file-hash-batch-size, value updated from 100 to 50.
  • io.gcs.max-batch-size, value stays the same at 100.
  • io.gcs.max-batch-duration, value stays the same at 5 seconds.

• New configuration parameters which should not require updating:

  • io.command-backpressure-staleness
  • io.backpressure-extension-log-threshold
  • load-control.io-normal-window-minimum
  • load-control.io-normal-window-maximum

• io.nio.parallelism was previously misspelled in reference.conf but not in Cromwell's configuration reading code. Only correct spellings of this configuration key had or will have effect.

66

66 Release Notes

Google Artifact Registry Support

Cromwell now supports call caching when using Docker images hosted on
Google Artifact Registry.

Google Image Repository Hashing Updates

The previously documented docker.hash-lookup.gcr configuration has been renamed to docker.hash-lookup.google and
now applies to both Google Container Registry (GCR) and Google Artifact Registry (GAR) repositories.
Support for the docker.hash-lookup.gcr-api-queries-per-100-seconds configuration key has been formally discontinued
and a bug preventing correct handling of docker.hash-lookup...throttle configuration has been fixed.
Please see Cromwell's bundled
reference.conf
for more details.

65

65 Release Notes

• An additional set of metrics relating to metadata age were added.

64

64 Release Notes

Intel Cascade Lake support on PAPI v2

On the PAPI v2 backends "Intel Cascade Lake" is now supported as a CPU platform. More details can be found
here.

63

63 Release Notes

Removed refresh token authentication mode

Google Pipelines API v1 supported authentication with refresh tokens, while v2 of the API does not.

Now that v1 has been discontinued and shut down, this version of Cromwell removes support for refresh tokens.

62

62 Release Notes

Downloading Access URLs

Added experimental support to download data during Google Cloud Life Sciences
jobs using DRS AccessURLs.

61

61 Release Notes

No labels update for Archived workflows

If - and ONLY if - you have metadata archiving turned on, then for a workflow whose metadata has been archived by Cromwell
according to the lifecycle policy, Cromwell will no longer add new labels or update existing labels for this workflow
coming through PATCH /labels endpoint.

60

60 Release Notes

Java 11

As of this version, a distribution of Java 11 is required to run Cromwell. Cromwell is developed, tested, and
containerized using AdoptOpenJDK 11 HotSpot.

Hybrid metadata storage ("carboniting") removed

Carboniting functionality has been removed from Cromwell.
There will be no effect for customers who store metadata permanently in the relational database (most common),
and there will also be no effect for customers who use the in-memory database.

Breaking change only for customers who explicitly enabled carbonite-metadata-service in their configuration to split
metadata storage between a relational database and Google Cloud Storage. If you had previously enabled carboniting and
deletion, any workflows marked as ArchivedAndPurged in your database will no longer be accessible via the Cromwell metadata API.