New or Improved Features¶
Version 14.0.x¶
Removed Features¶
Support for Amazon DocumentDB versions below 5.0 has been discontinued. Amazon DocumentDB 4.0 reports MongoDB wire protocol version 7, while current MongoDB drivers require wire protocol version 8, which corresponds to MongoDB 4.2 or later. If your Cluster Manager uses a version of Amazon DocumentDB below 5.0, you must upgrade your cluster to Amazon DocumentDB 5.0 or later before upgrading the Cluster Manager. Amazon Web Services supports in-place major version upgrades from the AWS Management Console or the AWS CLI.
Support for Azure Cosmos DB for MongoDB versions below 7.0 has been discontinued. If your Cluster Manager uses an older Azure Cosmos DB version, you must upgrade your cluster to version 7.0 or later before upgrading the Cluster Manager. See Microsoft’s Azure Cosmos DB for MongoDB upgrade guide for instructions.
Global Updates¶
Gurobi Compute Server now supports Gurobi Optimizer 14.0.x, in addition to past releases 12.0.x and 13.0.x. Support for Gurobi Optimizer 11.0.x has been discontinued; it is no longer supported nor packaged in the compute server.
Cluster Manager¶
The Cluster Manager Web UI has been rewritten in TypeScript. This brings stronger type safety across the frontend codebase, catching a broad class of errors at build time rather than at runtime, and sets up the UI for faster, safer feature development going forward. The rewrite preserves existing functionality and does not require any action from users.
The interactive API reference in the Cluster Manager Web UI, previously rendered with Swagger UI, now uses Scalar. This provides a more modern and easier-to-use interface for browsing and trying out the REST API, and reduces the size of the Web UI bundle. The API itself is unchanged and no action is required from users.
The Cluster Manager now decides if a Compute Server node is alive based only on its own clock. Previously,
a node whose clock was more than one minute behind the Cluster Manager’s was removed from the cluster at
every keepalive. Now, a difference between clocks no longer affects whether a node is considered alive: a
node is removed only after DISCONNECT_TIMEOUT without keepalive. Clock differences above 10 seconds are
logged and shown in a new SKEW column of grbcluster nodes.
New DISCONNECT_TIMEOUT (default 60) and REGISTRY_REFRESH_INTERVAL (default 20) properties, in
seconds, in grb_rsm.cnf.