v6.0.1

• Fix #423 : TaskRepositoy.findByChainDealIdAndTaskIndex() returns non-unique result

v6.0.0

What's new?

• #411 Added Prometheus actuators
• #413 Moved TEE workflow configuration to dedicated service

v5.1.1

What is patched?

• Updated management port and actuator endpoints

v5.1.0

What's new?

Fix WebSockets problem

#388 Enhance different task stages detection. When detecting unnotified contribute/reveal
we use the initialization block instead of the current block to lookup for the contribute/reveal metadata.
#386 Use a dedicated TaskScheduler for STOMP WebSocket heartbeats.
#386 Use a dedicated TaskScheduler for @Scheduled tasks.
#386 Use a dedicated TaskExecutor for @Async tasks.
#386 TaskService is now the entry point to update tasks.
#385 feature/task-replay

Also:

#396 Use the deal block as a landmark for a task.
#395 Keep the computing task list consistent.
#394 Worker should be instructed to contribute whenever it is possible (e.g. app/data download failure).
#393 Enhance worker lost detection.
#392 Enhance final deadline detection for tasks.

v5.0.0

What's new?

• #371 Dapp developers can browse worker computing logs over iexec-core API
• #374 Task result link is standardized and generic. It supports Ethereum, IPFS & Dropbox "storage" providers.
• #361 Result storage feature is moved to a dedicated component - https://github.com/iExecBlockchainComputing/iexec-result-proxy
• Full compatibility with new iExec Secret Managment Service - https://github.com/iExecBlockchainComputing/iexec-sms
• #370 Compatibility with latest PoCo 5.1.0 smart contracts

v4.0.1

What's fixed?

• More resistance to unsync Ethereum nodes #357

v4.0.0

What's new?

• Native-token sidechain compatibility #335
• GPU workers support #341
• Log aggregation #334

What's fixed?

• Database indexes #340
• JWT/challenge validity duration #339
• Worker freed after contribution timeout #338

v3.2.0

What is new?

• Bag Of Tasks (BoT): Bag Of Tasks Deals can now be processed by the middleware.
• Use of iexec input files: Sometimes external resources are needed to run a computation without using a dataset, that is what input files are for.
• Whitelisting: Now the core can define a whitelist of workers that are allowed to connect to the pool.
• Https: Workers can connect to the core using https.

What is patched?

• The project has been updated to java 11 and web3j 4.3.0
• Internal refactoring to handle replicates update better
• Limit workers that ask for replicates too often
• Update workers configuration when they disconnect/reconnect

v3.1.0

What's new?

• Full end-to-end encryption inside a Trusted Execution Environment (TEE) powered by Intel(R) SGX #307
• Implemented the Proof-of-Contribution (PoCo) Sarmenta's formula for a greater task dispatching #305

What's patched?

• Reopen task worflow is back #302
• A single FAILED replicate status when a completion is impossible #295
• WORKER_LOST is not set for replicates which are already FAILED #295 #299
• Restart when ethereum node is not available at start-up

v3.0.1

This release is a patch to fix an issue following the release of version 3.0.0.

When asking for a replicate, a worker sends the latest available block number from the node it is is connected to. If that node is a little bit behind the node the core is connected to, the worker will have a disadvantage compare to workers connected to more up-to-date nodes. To avoid this disadvantage, now the core waits for a few blocks before sending replicates to workers.