[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: Much more evolved database preparation script
- Subject: Re: Much more evolved database preparation script
- From: "Adam Thompson" <..hidden..>
- Date: Thu, 26 May 2011 16:51:19 -0500
I realize it's a bit late to jump in with this comment, but I think it has
to be made anyway:
In 2011, I only install software two ways:
1. via a distribution-native mechanism (i.e. pkginstall, yum/rpm,
apt/dpkg/deb, ports [à la *BSD])
2. manually (which includes "make install" - see below)
As a long-term UNIX admin, install scripts are anæthema - no vendor ever
has the same notion of where things should go that I do, or that my
distribution does. Install scripts that attempt to be multi-platform
invariably get something wrong because the author isn't intimately
familiar with each and every OS and version of said OS that the script
might be run on.
I realize the LSMB install procedure is complex enough that very few users
will want to attempt a purely-manual install, and therefore *something* is
needed. For standalone, (usually) compiled, software, the traditional
"tar...; ./configure...; make install" routine is acceptable, because all
"make install" is doing is copying some files around; the autoconf stage
first ensured all the dependencies were met and set up the makefile the
way I wanted it.
Getting an autoconf project off the ground in the first place is a major
project in and of itself - so I'm DEFINITELY NOT volunteering! This is
typical of highly mature, well-staffed GNU projects.
The "easy" approach is to pick a single target distro, and develop a
'port' or a package script that works flawlessly on that version of that
distro. Then let interested parties produce native packaging for their
favourite distro. It occurs to me that many (most?) people won't want to
run an accounting system on the same system as their fileserver, or their
webserver, or... so an appliance-based approach is viable. VM technology
is widespread enough that saying "here's the VM of CentOS/Debian/whatever
with LSMB preinstalled" has become an acceptable packaging technique. I'm
starting to buy "black-box" VMs from some of my vendors instead of
hardware appliances.
Those who don't like the appliance approach probably skew more towards #2
(manual install) anyway.
While I can't claim to represent everyone, nor even the majority of people
currently interested in the project, I believe my standpoint is valid for
the broader public who - largely - doesn't care how they get their
software working, just that it works.
So from my perspective, documenting the install process is a LOT more
helpful than an install script that doesn't quite work the way I want it
to. When I looked at using LSMB 1.3 over a year ago, I gave up because
the documentation on installing it was self-contradictory and had many
gaps; there was no way to do the install without using one of (IIRC) three
different scripts, none of which quite worked and all of which took
completely different approaches.
So, it's just my $0.02, but stop trying to accommodate everyone's personal
preferences, and pick a platform. *Then*, once it's working *somewhere*,
worry about getting it working everywhere.
-Adam Thompson
..hidden..
> -----Original Message-----
> From: Erik Huelsmann [mailto:..hidden..]
> Sent: Thursday, May 26, 2011 16:27
> To: Development discussion for LedgerSMB
> Subject: [Ledger-smb-devel] Much more evolved database preparation
> script
>
> Hi Luke,
>
>
> On Thu, May 26, 2011 at 1:13 AM, Luke <..hidden..>
> wrote:
> > Three additional things:
> >
> > 1. The two lines:
> >
> > psql_args=-h $HOST --password $PASS -U ledgersmb psql_cmd=psql
> > $psql_args -d $company_name
> >
> > Are not likely to work as expected. You probably want something like:
> >
> > psql_args="-h $HOST --password $PASS -U ledgersmb"
> > psql_cmd="psql $psql_args -d $company_name"
> >
> > 2. Where is the script getting $company_name?
> >
> > 3. Shouldn't "ledgersmb" be variablized?
>
> Thanks for your comments. The script has evolved much more now (still
> not tested though).
>
> The script isn't really an install script, because it doesn't copy files
> into
> place. Rather, it prepares the database for first operation by loading
> the
> right tables, functions, triggers, indices and what not.
>
> The script should help people to get their database in working order
> with
> much less hassle than the hassle I went through when I did the same in
> October last year.
>
> More comments very welcome!
>
>
> Bye,
>
>
> Erik.
>
> =========================
> #!/bin/sh
>
> # Script to explain the steps to take when installing LedgerSMB
>
> # Default variable values section
>
> owner=ledgersmb
> pass=LEDGERSMBINITIAL
> host=localhost
> port=5432
> srcdir=$PWD
> dstdir=$PWD
> coa=us/General.sql
> gifi=
> ADMIN_FIRSTNAME='Default'
> ADMIN_MIDDLENAME=NULL
> ADMIN_LASTNAME='Admin'
> ADMIN_USERNAME='admin'
> ADMIN_PASSWORD='admin'
>
>
> # Usage explanation section
>
>
> usage () {
> script_name=`basename $0`
> cat <<USAGE
> usage: $script_name [option1 .. optionN]
>
> This script wants to be run as the root user. If you don't, you'll be
> asked
> to enter the password of the 'postres' user
>
> This script creates a 'ledgersmb' user on the specified PostgreSQL host,
> if
> it does not exist. Then it proceeds to load the LedgerSMB database
> structure and content, loading Chart of Accounts data and GIFI data as
> specified in the argument list.
>
> After the database has been created, the script inserts a default user
> '$ADMIN_USER' (password: '$ADMIN_PASSWORD'), with these initial
> values:
>
> First name: $ADMIN_FIRSTNAME (NULL == none) Middle name:
> $ADMIN_MIDDLENAME (NULL == none)
> Last name: $ADMIN_LASTNAME (NULL == none)
> Country: 'US'
>
> This default user will be assigned all priviledges within the
> application.
>
> Available options:
> --srcdir The path where the sources for LedgerSMB are located
> [$srcdir]
> --dstdir The path where the sources will be located when invoked
> from the webserver [$dstdir]
> --host The PostgreSQL host to connect to (see 'man psql')
> [$host]
> --port The PostgreSQL port to connect to (see 'man psql')
> [$port]
> --company The name of the database to be created for the company
> [*]
> --owner The name of the superuser which is to become owner of
> the
> company's database [$owner]
> --password The password to be used to create the 'ledgersmb' user
> on the specified PostgreSQL server [$pass]
> --coa The path appended to SRCDIR/sql/coa/chart/ of the file
> to be
> used to load the Chart of Accounts data
> --gifi The path appended to SCRDIR/sql/coa/gifi/ of the file to
> be
> used to load the GIFI data with the chart of accounts
> --progress Echoes the commands executed by the script during setup
> --help Shows this help screen
>
> * These arguments don't have a default, but are required USAGE
>
>
> # Extract options and setup variables
> if ! options=$(getopt -l
> company,coa,gifi,srcdir,dstdir,password,host,port,help,progress --
> "$@")
> then
> exit 1
> fi
>
> set -- $options
> while test $# -gt 0 ;
> do
>
> shift_args=2
> case "$1" in
> --company)
> company_name=$2
> break;;
> --coa)
> coa=$2
> break;;
> --gifi)
> gifi=$2
> break;;
> --srcdir)
> srcdir=$2
> break;;
> --dstdir)
> dstdir=$2
> break;;
> --password)
> pass=$2
> break;;
> --host)
> host=$2
> break;;
> --port)
> port=$2
> break;;
> --progress)
> progress=yes
> shift_args=1
> break;;
> --help)
> usage
> exit 0;;
> esac
> shift $shift_args
> done
>
> if test -z "$company_name"
> then
> echo "missing or empty --company option"
> usage
> exit 1
> fi
>
> psql_args="-h $host --password $pass -U $owner"
> psql_cmd="psql $psql_args -d $company_name"
>
>
> if test -n "$progress"
> then
> # Use shell command-echoing to "report progress"
> set -x
> fi
>
> sed -e "s|WORKING_DIR|$dstdir|g" \
> $srcdir/ledgersmb-httpd.conf >$dstdir/ledgersmb.conf su -c
> "createuser -h $HOST --superuser -U postgres --password $PASS --no-
> inherit --encrypted ledgersmb" postgres
>
> createdb $psql_args -O $owner -E UNICODE $company_name createlang
> $psql_args plpgsql $company_name
>
>
> # Load the base file(s)
> # -- Basic database structure
> cat $srcdir/sql/Pg-database.sql | $psql_cmd # -- Additional database
> structure for module in `grep -v -E '^[[:space:]]*#'
> sql/modules/LOADORDER` do
> cat $srcdir/sql/modules/$module | $psql_cmd done # -- Authorizations
> sed -e "s/<\?lsmb dbname \?>/$company_name/g" \
> $srcdir/sql/modules/Roles.sql | $psql_cmd
>
>
> if test -n "$coa" ; then
> # Load a chart of accounts
> cat $srcdir/sql/coa/chart/$coa | $psql_cmd
> if test -n "$gifi" ; then
> cat $srcdir/sql/coa/gifi/$gifi | $psql_cmd
> fi
> fi
>
> cat <<CREATE_USER | $pgsql_cmd
> SELECT admin__save_user(NULL,
> person__save(NULL,
> 3,
> $ADMIN_FIRSTNAME,
> $ADMIN_MIDDLENAME,
> $ADMIN_LASTNAME,
> (SELECT id FROM country
> WHERE short_name = 'US')),
> $ADMIN_USERNAME,
> $ADMIN_PASSWORD);
>
> SELECT admin__add_user_to_role($ADMIN_USERNAME, rolname)
> FROM pg_roles WHERE rolname like 'lsmb_$company_name%';
>
> CREATE_USER
> =========================
>
> ------------------------------------------------------------------------------
> vRanger cuts backup time in half-while increasing security.
> With the market-leading solution for virtual backup and recovery, you
> get blazing-fast, flexible, and affordable data protection.
> Download your free trial now.
> http://p.sf.net/sfu/quest-d2dcopy1
> _______________________________________________
> Ledger-smb-devel mailing list
> ..hidden..
> https://lists.sourceforge.net/lists/listinfo/ledger-smb-devel