On Monday, November 16, 2015, Garrett Barton <[email protected]>
wrote:

> I have been using enqueueMutateBatch lately, it was much faster (even being
> out of band) than mutateBatch was for me.  I'm not wed to it, some slow
> down is alright.  What was wrong with its impl?


Yeah I know it's faster. However if a shard server fails while there are
mutates on the queue it will loose data. I was thinking of something like
allowing the row mutation to be pushed to a Kafka queue for persistence and
letting the indexers pull from there in bulk.  What do think?

Aaron



>
> On Mon, Nov 16, 2015 at 4:54 PM, Aaron McCurry <[email protected]
> <javascript:;>> wrote:
>
> > I would like to remove the following mutation methods due to the half
> baked
> > nature of their implementation.  Does anyone have any heartburn over
> this?
> >
> >
> >
> https://github.com/apache/incubator-blur/blob/master/scripts/interface/Blur.thrift#L1169-L1175
> >
> >
> https://github.com/apache/incubator-blur/blob/master/scripts/interface/Blur.thrift#L1185-L1231
> >
> > Aaron
> >
>

Reply via email to