drill-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Paul Rogers <prog...@maprtech.com>
Subject Re: Question re "obscure" Drill scripts
Date Wed, 22 Jun 2016 00:07:40 GMT
As a follow up…

I’d still appreciate hints for the “obscure” scripts.

I propose to make a change to the four sqlline scripts. Today, the sqlline scripts use the
same memory settings as the drillbit. This can lead to overly heavy client processes. The
change is:

The three client scripts (sqlline, drill-local, drill-conf) will use client-only configuration
options (especially for memory.)

The embedded script (drill-embedded) will continue use the Drillbit memory options (heavy
heap, direct memory and code cache.)

This change will be transparent to users.

Anything I’m missing? Do we need the heavy heap & direct memory when the Drill code
acts only as a client?

Thanks,

- Paul

> On Jun 21, 2016, at 11:48 AM, Paul Rogers <progers@maprtech.com> wrote:
> 
> Hi All,
> 
> The Drill-on-YARN (DoY) stuff will include some much-needed clean-up of the Drill scripts
in $DRILL_HOME/bin, so I need to retest all the scripts. Three scripts seem pretty obscure.
But, let’s start with the known ones.
> 
> The “primary” scripts seem to be:
> 
> drillbit.sh — User command to start/stop/monitor a Drill daemon
> drill-config.sh — Internal script to set up the Drill launch environment.
> runbit — Internal script to build up the Java command line for a drillbit and run it.
> sqlline — User script to launch SqlLine.
> 
> One question that immediately pops to mind: why do some scripts use “.sh” and some
not? One could argue that user-commands should omit the .sh as sqlline does. But drillbit.sh
uses .sh. And, the purely internal drill-config.sh and runbit made opposing choices… Just
an artifact, or is there some subtle pattern?
> 
> There are various wrappers around sqlline as described in the Apache Drill docs: https://drill.apache.org/docs/starting-drill-in-distributed-mode/
and https://drill.apache.org/docs/starting-drill-on-linux-and-mac-os-x/
> 
> drill-conf — Wrapper for sqlline, uses drill config to find Drill. Seems this one needs
fixing to use a config other than the hard-coded $DRILL_HOME/conf location.
> drill-embedded — Starts a drill “embedded” in SqlLine, using a local ZK.
> drill-localhost — Wrapper for sqlline, uses a local ZK.
> 
> Then there are three very obscure scripts. What do these do, and, more importantly, are
they still used?
> 
> drillbit — Appears to be a analog of drillbit.sh, but seems patterned for some system
with which I’m not familiar. It also assumes Drill is installed in /opt/drill. Any ideas?
> dumpcat — Runs an undocumented DumpCat Java class. The only mentions of this on Apache
Drill are two JIRA entries.
> submit-plan — Appears to directly submit a physical plan. Documented only in a JIRA:
https://issues.apache.org/jira/browse/DRILL-2107
> 
> Thanks,
> 
> - Paul
> 
> 


Mime
View raw message