[go: up one dir, main page]
Skip to content

Transaction manager

The transaction manager is responsible for ensuring that a write operation (mutator RPC) occurs successfully and consistently to each backend node. Some initial scenarios:

  • If a transaction fails due to an application level error, the transaction manager should rollback the transaction to the previous state and forward the primary replica's error.
  • If the transaction succeeds on a quorum of nodes, but there are failures on some, the failures must be reported back to the replication manager so that those nodes can be scheduled for repair (replication).
  • Likewise, if a transaction succeeds, but the checksums reported back at the end of the transaction differ between primary and secondary replicas, then all secondaries with irregular checksums must be reported to the replication manager so that the irregularities can be rectified.
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information