Rodeos per block ack

[heifner]

Change Description

• Wait on ack from AMQP before continuing on to processing the next block.

Change Type

Select ONE:

• Documentation

• Stability bug fix

• Other

• Other - special case

Testing Changes

Select ANY that apply:

• New Tests

• Existing Tests

• Test Framework

• CI System

• Other

Consensus Changes

• Consensus Changes

API Changes

• API Changes

Documentation Additions

• Documentation Additions

[swatanabe-b1]

I don't understand the motivation for this change.

• commitTransaction only guarantees that rabbitmq has taken responsibility for the message. It does not guarantee that the message has been delivered.
• Rodeos can still be ahead of what has been successfully published to rabbit, so it doesn't allow you to get rid of the local storage for messages.

[heifner]

1. commitTransaction only guarantees that rabbitmq has taken responsibility for the message. It does not guarantee that the message has been delivered.
2. Rodeos can still be ahead of what has been successfully published to rabbit, so it doesn't allow you to get rid of the local storage for messages.

1. The desired responsibility is that rabbitmq has taken responsibility of the message. Once it has ack'ed the message it is rabbitmq responsibility to deliver or recover the message on DR.
2. Rodeos commits the changes to rocksdb after the ack from rabbitmq.

[swatanabe-b1]

2. Rodeos commits the changes to rocksdb after the ack from rabbitmq.

That's true, but the commit happens regardless of whether the ack was received or not. The only thing that I see that checks for app().quit() is write_deltas, which happens before the filter is processed. rodeos_db_snapshot::end_block may commit the changes before we return to appbase and check the quitting flag.

[heifner]

1. Rodeos commits the changes to rocksdb after the ack from rabbitmq.

That's true, but the commit happens regardless of whether the ack was received or not. The only thing that I see that checks for app().quit() is write_deltas, which happens before the filter is processed. rodeos_db_snapshot::end_block may commit the changes before we return to appbase and check the quitting flag.

Oh, I see what you mean now, thanks.