AnsweredAssumed Answered

3.0.2 MFS not respecting warden.config

Question asked by snelson on Oct 21, 2014
Latest reply on Oct 31, 2014 by snelson
I'm trying to allocate more memory to MFS because our use case is heavy on M7 tables. We do 30-60 scans/sec and when load is heavy, it cripples performance. I don't see any other bottlenecks that could be affecting performance when I look at our Ganglia graphs.

[This doc][1] indicates I can change the amount of memory allocated to MFS by using the service.command.mfs.heapsize.* configuration in warden.config. However when I adjust the memory and restart warden, it doesn't change the amount of memory used by the process. When I look at the process in `ps`, it's always

    /opt/mapr/server/mfs -b -p 5660 -m 5245 -g -O /opt/mapr/conf/mapr-clusters.conf -V

This indicates to me that 5245MB is allocated to the process. This is only ~35% of the total memory of my system. According to the same doc, 35% is the default amount of memory allocated to MFS (when running M7 tables). Are the configurations in warden.config ignored completely? Does my version not support it (I'm running 3.0.2)? How do I increase the amount of memory that MFS is allowed to use?


  [1]: http://doc.mapr.com/display/MapR/Memory+Allocation+for+Nodes

Outcomes