List:Cluster« Previous MessageNext Message »
From:Wasif Riaz Malik Date:June 21 2012 6:20pm
Subject:Re: ClusterJ distribution awareness
View as plain text  

On Thu, Jun 21, 2012 at 5:11 PM, Bernd Ocklin <bernd.ocklin@stripped>wrote:

> Hi,
> Am 18.06.2012 14:10, schrieb Wasif Riaz Malik:
>  Hi,
>> I am using ClusterJ 7.1.15a to query a cluster of ten nodes. I have
>> noticed
>> that the transactions are sent only to three nodes (even though the data
>> is
>> evenly spread out on all ten nodes).
> can you highlight what exactly you are seeing? Do you see 3 nodes being
> more CPU loaded? Do you take stats? Which stats do you base this on? ...

I was monitoring the stats using ndb_mgm (ALL CLUSTERLOG STATISTICS=15). It
seemed as if the queries were being sent to only three nodes because only
they had *Trans count *greater than zero. I was assuming that each ordered
index scan sent by ClusterJ affects the *Trans count *value.* *But I have
found out from the docs (and some testing) that ordered index scans only
show up as *Range scans* in the cluster logs and not in the *Trans count*.

The range scans are getting evenly distributed (in round robin) on all ten

> Bernd
>> How does clusterj decide the transaction coordinator be default (if no
>> hints are given)?  Is it possible to distribute the transactions in a
>> round
>> robin fashion (without passing hints for each transaction)?
>> Regards,
>> Wasif Riaz Malik
> --
> MySQL Cluster Mailing List
> For list archives:
> To unsubscribe:

ClusterJ distribution awarenessWasif Riaz Malik18 Jun
  • RE: ClusterJ distribution awarenessAndrew Morgan18 Jun
    • Re: ClusterJ distribution awarenessCraig L Russell18 Jun
  • Re: ClusterJ distribution awarenessBernd Ocklin21 Jun
    • Re: ClusterJ distribution awarenessWasif Riaz Malik21 Jun