Arvados 2.3.0 Release Notes
October 5, 2021
The Arvados team is pleased to announce Arvados 2.3.0. This is a major upgrade, with many new features as well as bugfixes. We recommend that new and existing installations of 2.2.2 or earlier upgrade to 2.3.0. See Upgrading Arvados for upgrade instructions.
Initial LSF support
This new release adds support for dispatching containers to Spectrum LSF clusters, allowing the integration of Arvados to additional HPC environments.
For more information please check our documentation page.
Initial Singularity support
This major release begins supporting running containers with the Singularity Runtime.
We’ve tested this version against Singularity 3.7.4, and its support should be considered in alpha state. For more information please check our documentation page.
Improved Storage Classes support
We’ve completed a feature that enables better data management. With storage classes, the site administrator can optimize/customize different storage backends where to put the data blocks depending on the organization’s policies.
The new features include the ability of the different SDKs to directly upload the data to specific storage class(es), saving time and bandwidth costs. On the backend side, collection listings can now be filtered by storage classes.
The administrator can configure the cluster to set multiple default classes to data collections when the client doesn’t specify them. Also, different applications and workflow outputs can have their storage class(es) set. You can learn more about this on the user’s guide.
Regarding bandwidth & time savings, the keepstore services parallelize data write operations when dealing with multiple storage classes that correspond to different volumes and prioritize read operations on storage classes managed by multiple volumes. Please refer to the admin documentation for more information.
Workbench2 now integrates the browser-based shell node access on its own instead of redirecting to the older Workbench1. This includes support for session auto-logout.
Workbench2 collection’s file browser
The new collection file browser allows navigating through more significant collections faster by loading partial file listings on demand instead of requesting the entire collection at once. This is most noticeable on collections with many directories and/or files.
Group-sync tool’s new feature
This release’s group syncing tool
arvados-sync-groups allows case insensitive
username matching as described on the admin guide.
Collection manifest handling
Collection manifest signing performance and caching are notably improved. This is specially helpful when dealing with really big data collections.
Removed deprecated full text search database indexes, allowing collections with lots of files and directories.
The pagination display now works correctly when showing empty tables.
Upload speed indication is now fixed showing
MB/s instead of
Data access fixes
WebDAV caching fix improves performance on huge collections. PySDK FUSE driver locking fix further increases its robustness.
Error reporting fixes
Request-ID propagation fixes improve error traceability and debugging. Keepproxy spurious 413 error fix avoids future debugging confusion.
Cached user records handling in a satellite cluster auto-resolves