Re: CommitBatchSize vs. CommitBatchThreshold for transactional dis

I strongly suspect this 1 off command is the commit command.

No bug here.

On Oct 28, 8:53 pm, "Gopal Ashok [MSFT]" <gopalas...@xxxxxxxxxxx>
wrote:
Just had a chat with Paul on this one

Commithreshold is honored at a transaction boundary. Let's take the
following example

Commitbatchsize 10
Commitbatchthreshold 20

Transaction1 has 19 commands
Transaction 2 has 11 commands
Transaction 3 has 5 commands

With the above setting, the distribution agent reader thread will set the
commit flag for the command block after we read all the commands in
transaction 2. This is because we check the commitbatch size and
commitbatchthreshold after a transaction boundary is reached. We dont issue
a commit after we read the first command (20th command in that batch) in
transaction 2. So this commit batch on the subscriber will have committed 2
transactions with 30 commands even though the threshold is 20.

So the batch will be committed either when we hit the commitbatchsize OR
when a transaction crosses the command threshold before the commitbatchsize
is hit.

With regard to the threshold adding one more command that seems to be a bug.

Hope this helps.

"Paul Ibison" <Paul.Ibi...@xxxxxxxxxxxxxxxxxxxxxxx(donotspam)> wrote in
messagenews:0A8275F6-3C9A-4388-AAF4-4DC7847E482C@xxxxxxxxxxxxxxxx

Definitely agree that there's some need for clarification (apart form the
bug
that is) and have raised up all this with MS to try to get a clearer
picture
so I can request BOL to be updated. Will update this thread when I know
more.
Many thanks,
Paul Ibison,www.replicationanswers.com

.

Relevant Pages

  • Re: [Info-Ingres] commit question
    ... The next block has been turned into a rollback ... hence the next command executed is the rollback and not the commit. ... The situation we had was that the "modify to truncate" command succeeded ...
    (comp.databases.ingres)
  • Timeout regression introduced by 242f9dcb8ba6f68fcd217a119a7648a4f69290e9
    ... Commit 242f9dcb8ba6f68fcd217a119a7648a4f69290e9 introduces a strange ... pointer from the issued command onto eh_cmd_q breaking libata EH ... ata2.00: XXX skipping completion ... ata2: hard resetting link ...
    (Linux-Kernel)
  • Re: Multiple concurrent isolated transactions on ADO Dataset
    ... When a command is called, the GetData is called, and the ... transaction which succeeds or fails. ... > on 89 - just in the small dataset that contains the seat reservation ... >> commit), whilst the unchanged version should be visible to all other ...
    (microsoft.public.dotnet.framework.adonet)
  • Re: [regression] CD-DA delay needed after insertion
    ... Inserting an audio CD ... Running the following command as soon as the blue CD/DVD/BD drive LED ... Reverting that commit doesn't fix the problem. ...
    (Linux-Kernel)
  • Re: XP x 3+clipper=SLOW
    ... they were causeb by the "commit" command. ... > (with mapping vs substing locally and mapping over the network). ... > Clipper applications and similar file sharing apps are run. ...
    (comp.lang.clipper)