cassandra-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From DuyHai Doan <doanduy...@gmail.com>
Subject Re: Writing to multiple tables
Date Mon, 16 Mar 2015 19:48:52 GMT
"Is BATCH the recommended way of updating all three tables at one go so
that the information between the three tables is consistent ? "

 If you're thinking about "atomicity", no it's not atomic. Indeed with
logged batches, what you gain is automatic retry done for you by the
coordinator in case of failure in the middle of the batch, and that's
pretty much it. The logged batch relieves the developer of the burden of
having to set up a retry strategy client side

 In general, it is recommended not to put too much data/statements in a
batch because the coordinator will block until having the ack for each
statement in the batch. Having thousands of statements in the same batch or
few statements with huge payload is definitely a bad idea.

"Are there alternate approaches and recommendations?"

If you don't care about managing retry yourself (or relying on the retry
policy of the driver), use executeAsync() to dispatch all statements over
the nodes of your cluster


On Mon, Mar 16, 2015 at 8:38 PM, Robert Coli <rcoli@eventbrite.com> wrote:

> On Mon, Mar 16, 2015 at 12:13 PM, Viswanathan Ramachandran <
> vish.ramachandran@gmail.com> wrote:
>
>> Is BATCH the recommended way of updating all three tables at one go so
>> that the information between the three tables is consistent ?
>>
>
> As a general statement :
>
> If you are looking to update multiple tables in a transaction, maybe
> Cassandra is not the ideal data-store for you. It is pretty un-Cassandric
> to want to do such a thing.
>
> =Rob
>
>

Mime
View raw message