You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@ignite.apache.org by Bob Krier <rk...@axway.com> on 2015/11/06 17:23:23 UTC

Disk Secondary FileSystem

Does anyone know if there is an implementation of IgfsSecondaryFileSystem that simply stores to old fashioned spinning platter disks, NFS, SMB, etc.  I’m doing development on a laptop and need to start and stop nodes and the entire cluster a lot.  Using Hadoop as a secondary FS requires lots of heap just to do anything or you get OutOfMemoryErrors.  

I would think an implementation of this would help many developers.  I’ve tried to make one work, but it seems a little over my head and the API docs are not exactly clear to me.



Re: Disk Secondary FileSystem

Posted by Denis Magda <dm...@gridgain.com>.
Hi,

Presently IGFS supports only Hadoop as a persistent storage out of the box.

As far as I know Hadoop shouldn't consume much of RAM and OutOfMemoryError
(OOM) you're getting most likely to happen because of IGFS that stores too
much data and there is no more physical memory available on your machine. So
switching to other distributed persistent storage like SMB or NFS shouldn't
help you.

As I suggested in the other thread you can try to use IGFS eviction policy
to get rid of OOM:
http://apache-ignite-users.70518.x6.nabble.com/OutOfMemoryError-with-Hadoop-backing-filesystem-td1854.html#a1885

Regards,
Denis



--
View this message in context: http://apache-ignite-users.70518.x6.nabble.com/Disk-Secondary-FileSystem-tp1867p1886.html
Sent from the Apache Ignite Users mailing list archive at Nabble.com.