Hi Jun, Thanks for the comments.
200.1, 200.2, 200.3 were all oversights which have been fixed. 201.1: This has been added to the BeginTxnRequest now. If the transaction.timeout.ms value is greater than max.transaction.timeout.ms, then the BeginTxnRequest will fail with a `InvalidTransactionTimeout` error. The document has been updated to reflect this. 201.2: I added configurations for number of partitions, number of replicas, and segment size. I don't think the compression configuration makes sense because each Message set in the transaction log will have exactly one message, and compression won't really buy us much. Do let me know if you disagree. 201.3: The default has been updated to `read_uncommitted`, which is the new name for `all`. On Mon, Feb 13, 2017 at 1:21 PM, Jun Rao <j...@confluent.io> wrote: > Hi, Guozhang, > > Thanks for the proposal. I made a pass of the wiki and had the following > comments. > > 200. Message format: > 200.1 MaxTimestampDelta: Does that need to be delta since it's always a > fixed size in64? > 200.2 The wiki says "At the end we still maintains a message-level CRC". Is > that still valid? > 200.3 In the ProducerRequest, do we need messageSet size? > 200.4 One of the things that we may want to add in the future is KIP-82 > (per record header). It would be useful to think a bit how easy it is to > support that with the new message format. > > > 201. Configurations: > 201.1 transaction.timeout.ms in the producer: It seems that it's missing > in > BeginTxnRequest? Also, what happens when the value is larger than > max.transaction.timeout.ms on the broker? > 201.2 For the internal transactional topic, do we need additional broker > side configurations to control # of partitions, # of replicas, compression > codec, segment size like the offset topic? > 201.3 isolation.level: It says the default is "all", but there is no option > for "all". > > Thanks, > > Jun > > On Wed, Feb 1, 2017 at 8:13 PM, Guozhang Wang <wangg...@gmail.com> wrote: > > > Hi all, > > > > We would like to start the voting process for KIP-98. The KIP can be > found > > at > > > > https://cwiki.apache.org/confluence/display/KAFKA/KIP- > > 98+-+Exactly+Once+Delivery+and+Transactional+Messaging > > > > Discussion thread can be found here: > > > > http://search-hadoop.com/m/Kafka/uyzND1jwZrr7HRHf?subj=+ > > DISCUSS+KIP+98+Exactly+Once+Delivery+and+Transactional+Messaging > > > > Thanks, > > > > -- > > -- Guozhang > > >