Thanks for reply I'm using HAWQ 2.0.1.0 installed on single host ( 40 cpu , over 100Gigs of RAM). I wonder how to configure single host HAWQ instance to make use of host resources. At this moment im using HAWQ without yarn ,when i set hawq_rm_stmt_nvseg=X; only 6 of X virtual segments spawned by HAWQ use 100% cpu, rest virtual segments are idle
my rm settings show hawq_rm_memory_limit_perseg -> 128GB show hawq_rm_nvcore_limit_perseg -> 40 show default_hash_table_bucket_number -> 12 2016-09-29 15:44 GMT+02:00 Vineet Goel <[email protected]>: > Dominik, > > What version are you using? If latest version, hawq expand (gpexpand) is > not required and cluster can be expanded dynamically with just a few steps > very easily. > > If using Ambari, see: http://hdb.docs.pivotal. > io/20/admin/ambari-admin.html#amb-expand > > Otherwise, see: http://hdb.docs.pivotal.io/20/admin/ClusterExpansion.html > > - Vineet > > > On Thu, Sep 29, 2016 at 8:12 AM Dominik Choma <[email protected]> > wrote: > >> Hi all, >> >> I want to increase parallelism level and hardware utilization by >> increasing number of segments per host. In PHD it can be done via gpexpand >> http://pivotalhd-210.docs.pivotal.io/doc/2010/ >> ExpandingtheHAWQSystem.html#ExpandingtheHAWQSystem- >> IncreasingSegmentsPerHost. >> Does HAWQ have some utility similar to gpexpand ? >> >> Dominik >> >>
