Hi,

Is there a rule of thumb for the maximum number of updates before issuing a
commit?

For example, I'm using a MapReduce job for indexing a table in HBase... but
the problem is that I can't just let the reducers commit whenever they want
or else commits tend to happen at the same time and therefore resulting in
exceeding the maximum number of searchers warmed at the same time. On the
other hand, I can't imagine updating let's say 2 000 000 records in the
index in a distributed manner and issuing a commit only at the end...

Sebastien

Reply via email to