Re: MapReduce with Apache-Ignite

2016-10-27 Thread vkulichenko
lalit wrote
> You've mentioned that spilling task is worked upon. Does it mean in the
> current version because of eviction we may loose mapper output? or the
> node will crash complaining the out of memory error?

In this case Ignite will consume all the available memory and crash with
OOME.

-Val



--
View this message in context: 
http://apache-ignite-users.70518.x6.nabble.com/MapReduce-with-Apache-Ignite-tp8007p8560.html
Sent from the Apache Ignite Users mailing list archive at Nabble.com.


Re: MapReduce with Apache-Ignite

2016-10-14 Thread Vladimir Ozerov
Hi,

Currently mapper output is stored in-memory only. We are working on spilling
algorithm at the moment which will flush part of it to the disk according to
some pre-configured thresholds.

As per job logs, they are not supported at the moment by Ignite in the same
way as it is done for Hadoop. We add similar functionality soon. 



--
View this message in context: 
http://apache-ignite-users.70518.x6.nabble.com/MapReduce-with-Apache-Ignite-tp8007p8314.html
Sent from the Apache Ignite Users mailing list archive at Nabble.com.


Re: MapReduce with Apache-Ignite

2016-09-29 Thread lalit
Regarding first point about mapper output - I intend to know that the mapper
output will not be in-memory. Thus I/O operation on the local filesystem at
mapper side still exists?

Regarding second point - I tried running the job using the hadoop
accelerator, I want to locate the logs for the job id, the mapper tasks &
reducer tasks.



--
View this message in context: 
http://apache-ignite-users.70518.x6.nabble.com/MapReduce-with-Apache-Ignite-tp8007p8018.html
Sent from the Apache Ignite Users mailing list archive at Nabble.com.


Re: MapReduce with Apache-Ignite

2016-09-29 Thread Vladimir Ozerov
Hi,

1. Mapper output should be written to the same place as if it job was run
through native Apache Hadoop engine. Apache Ignite Hadoop Accelerator can
work without IGFS at all.
2. Currently you will not see jobs in Resource Manager because they are
executed through separate engine. We will improve this in future. Please
clarify what logs do you mean?

Vladimir.



--
View this message in context: 
http://apache-ignite-users.70518.x6.nabble.com/MapReduce-with-Apache-Ignite-tp8007p8016.html
Sent from the Apache Ignite Users mailing list archive at Nabble.com.