spark-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Du Li <>
Subject Re: how to use rdd.countApprox
Date Wed, 13 May 2015 19:13:49 GMT
Actually I tried that before asking. However, it killed the spark context. :-)

     On Wednesday, May 13, 2015 12:02 PM, Tathagata Das <> wrote:

 That is a good question. I dont see a direct way to do that. 
You could do try the following 
val jobGroupId = <group-id-based-on-current-time>rdd.sparkContext.setJobGroup(jobGroupId)val
approxCount = rdd.countApprox().getInitialValue   // job launched with the set job grouprdd.sparkContext.cancelJobGroup(jobGroupId)
          // cancel the job

On Wed, May 13, 2015 at 11:24 AM, Du Li <> wrote:

Hi TD,
Do you know how to cancel the rdd.countApprox(5000) tasks after the timeout? Otherwise it
keeps running until completion, producing results not used but consuming resources.

     On Wednesday, May 13, 2015 10:33 AM, Du Li <> wrote:

  Hi TD,
Thanks a lot. rdd.countApprox(5000).initialValue worked! Now my streaming app is standing
a much better chance to complete processing each batch within the batch interval.

     On Tuesday, May 12, 2015 10:31 PM, Tathagata Das <> wrote:

 From the code it seems that as soon as the " rdd.countApprox(5000)" returns, you can call
"pResult.initialValue()" to get the approximate count at that point of time (that is after
timeout). Calling "pResult.getFinalValue()" will further block until the job is over, and
give the final correct values that you would have received by "rdd.count()"
On Tue, May 12, 2015 at 5:03 PM, Du Li <> wrote:

I tested the following in my streaming app and hoped to get an approximate count within 5
seconds. However, rdd.countApprox(5000).getFinalValue() seemed to always return after it finishes
completely, just like rdd.count(), which often exceeded 5 seconds. The values for low, mean,
and high were the same.
val pResult = rdd.countApprox(5000)val bDouble = pResult.getFinalValue()logInfo(s"countApprox().getFinalValue():
low=${bDouble.low.toLong}, mean=${bDouble.mean.toLong}, high=${bDouble.high.toLong}")
Can any expert here help explain the right way of usage?


     On Wednesday, May 6, 2015 7:55 AM, Du Li <> wrote:

 I have to count RDD's in a spark streaming app. When data goes large, count() becomes expensive.
Did anybody have experience using countApprox()? How accurate/reliable is it? 
The documentation is pretty modest. Suppose the timeout parameter is in milliseconds. Can
I retrieve the count value by calling getFinalValue()? Does it block and return only after
the timeout? Or do I need to define onComplete/onFail handlers to extract count value from
the partial results?




View raw message