zc.recipe.rhrc 1.4.2

ZC Buildout recipe for Redhat RC scripts

This package provides a zc.buildout recipe for creating Red-Hat Linux compatible run-control scripts.

Contents

• Changes

  • 1.4.2 (2012-12-20)

  • 1.4.1 (2012-08-31)

  • 1.4.0 (2012-05-18)

  • 1.3.0 (2010/05/26)

    • New Features

    • Bugs Fixed

  • 1.2.0 (2009/04/06)

  • 1.1.0 (2008/02/01)

  • 1.0.0 (2008/01/15)

• Detailed Documentation

  • Create Red-Hat Linux (chkconfig) rc scripts

    • Working with existing control scripts

    • Multiple processes

    • Independent processes

    • Deployments

  • Process Management

  • Regression Tests

• Download

Changes

1.4.2 (2012-12-20)

Fixed: Errors were raised if stopping a run script failed during

uninstall. This could cause a buildout to be wedged, because you couldn’t uninstall a broken/missing run script.

1.4.1 (2012-08-31)

Fixed: Processes weren’t started on update.

In a perfect world, this wouldn’t be necessary, as in the update case, the process would already be running, however, it’s helpful to make sure the process is running by trying to start it.

1.4.0 (2012-05-18)

• Added optional process-management support. If requested, then run scripts are run as part of install and uninstall.

• Fixed: missing test dependency on zope.testing

1.3.0 (2010/05/26)

New Features

• A new independent-processes option causes multiple processes to be restarted independently, rather then stoping all of the and the starting all of them.

Bugs Fixed

• Generated run scripts had trailing whitespace.

1.2.0 (2009/04/06)

displays the name of the script being run for each script when it is started, stopped, or restarted

1.1.0 (2008/02/01)

Use the deployment name option (as provided by zc.recipe.deployment 0.6.0 and later) if present when generating script names.

Use the deployment rc-directory as the destination when a deployment is used.

Use /sbin/chkconfig rather than chkconfig, as I’m told it is always in that location and rarely in anyone’s path. :)

1.0.0 (2008/01/15)

Initial public release

Detailed Documentation

Create Red-Hat Linux (chkconfig) rc scripts

The zc.recipes.rhrc recipe creates Red-Hat Linux (chkconfig) rc scripts. It can create individual rc scripts, as well as combined rc scripts that start multiple applications.

The recipe has a parts option that takes the names of sections that define run scripts. They should either:

• Define a run-script option that contains a one-line shell script, or

• The file /etc/init.d/PART should exist, where PART is the part name.

A simple example will, hopefully make this clearer.

>>> demo = tmpdir('demo')

>>> write('buildout.cfg',
... """
... [buildout]
... parts = zoperc
...
... [zoperc]
... recipe = zc.recipe.rhrc
... parts = zope
... dest = %(dest)s
...
... [zope]
... run-script = /opt/zope/bin/zopectl -C /etc/zope.conf
... """ % dict(dest=demo))

Normally the recipe writes scripts to /etc/init.d. We can override the destination, which we’ve done here, using a demonstration directory. We specified a that it should get run-script source from the zope section. Here the zope section is simply a configuration section with a run-script option set directly, but it could have been a part with a run-script option computed from the recipe.

If we run the buildout:

>>> print system('bin/buildout'),
Installing zoperc.

We’ll get a zoperc script in our demo directory:

>>> ls(demo)
-  zoperc

>>> cat(demo, 'zoperc')
#!/bin/sh
<BLANKLINE>
# This script is for adminstrator convenience.  It should
# NOT be installed as a system startup script!
<BLANKLINE>
<BLANKLINE>
case $1 in
  stop)
<BLANKLINE>
    /opt/zope/bin/zopectl -C /etc/zope.conf $*
<BLANKLINE>
    ;;
  restart)
<BLANKLINE>
    ${0} stop
    sleep 1
    ${0} start
<BLANKLINE>
    ;;
  *)
<BLANKLINE>
    /opt/zope/bin/zopectl -C /etc/zope.conf $*
<BLANKLINE>
    ;;
esac
<BLANKLINE>

There are a couple of things to note about the generated script:

• It uses $* to pass arguments, so arguments can’t be quoted. This is OK because the arguments will be simple verbs like start and stop.

• It includes a comment saying that the script shouldn’t be used as a system startup script.

For the script to be used for system startup, we need to specify run-level information. We can to that using the chkconfig option:

>>> write('buildout.cfg',
... """
... [buildout]
... parts = zoperc
...
... [zoperc]
... recipe = zc.recipe.rhrc
... parts = zope
... dest = %(dest)s
... chkconfig = 345 90 10
... chkconfigcommand = echo
...
... [zope]
... run-script = /opt/zope/bin/zopectl -C /etc/zope.conf
... """ % dict(dest=demo))

Here we included a chkconfig option saying that Zope should be started at run levels 3, 4, and 5 and that it’s start and stop ordered should be 90 and 10.

For demonstration purposes, we don’t really want to run chkconfig, so we use the chkconfigcommand option to tell the recipe to run echo instead.

>>> print system('bin/buildout'),
Uninstalling zoperc.
Running uninstall recipe.
Installing zoperc.
--add zoperc

Now the script contains a chkconfig comment:

>>> cat(demo, 'zoperc')
#!/bin/sh
<BLANKLINE>
# the next line is for chkconfig
# chkconfig: 345 90 10
# description: please, please work
<BLANKLINE>
<BLANKLINE>
case $1 in
  stop)
<BLANKLINE>
    /opt/zope/bin/zopectl -C /etc/zope.conf $* \
      </dev/null
<BLANKLINE>
    ;;
  restart)
<BLANKLINE>
    ${0} stop
    sleep 1
    ${0} start
<BLANKLINE>
    ;;
  *)
<BLANKLINE>
    /opt/zope/bin/zopectl -C /etc/zope.conf $* \
      </dev/null
<BLANKLINE>
    ;;
esac
<BLANKLINE>

We can specify a user that the script should be run as:

>>> write('buildout.cfg',
... """
... [buildout]
... parts = zoperc
...
... [zoperc]
... recipe = zc.recipe.rhrc
... parts = zope
... dest = %(dest)s
... chkconfig = 345 90 10
... chkconfigcommand = echo
... user = zope
...
... [zope]
... run-script = /opt/zope/bin/zopectl -C /etc/zope.conf
... """ % dict(dest=demo))

>>> print system('bin/buildout'),
Uninstalling zoperc.
Running uninstall recipe.
--del zoperc
Installing zoperc.
--add zoperc

Note the –del output. If we hadn’t set the chkconfigcommand to echo, then chkconfig –del would have been run on the zoperc script.

>>> cat(demo, 'zoperc')
#!/bin/sh
<BLANKLINE>
# the next line is for chkconfig
# chkconfig: 345 90 10
# description: please, please work
<BLANKLINE>
<BLANKLINE>
if [ $(whoami) != "root" ]; then
  echo "You must be root."
  exit 1
fi
<BLANKLINE>
case $1 in
  stop)
<BLANKLINE>
    su zope -c \
      "/opt/zope/bin/zopectl -C /etc/zope.conf $*" \
      </dev/null
<BLANKLINE>
    ;;
  restart)
<BLANKLINE>
    ${0} stop
    sleep 1
    ${0} start
<BLANKLINE>
    ;;
  *)
<BLANKLINE>
    su zope -c \
      "/opt/zope/bin/zopectl -C /etc/zope.conf $*" \
      </dev/null
<BLANKLINE>
    ;;
esac
<BLANKLINE>

Note that now the su command is used to run the script. Because the script is included in double quotes, it can’t contain double quotes. (The recipe makes no attempt to escape double quotes.)

Also note that now the script must be run as root, so the generated script checks that root is running it.

If we say the user is root:

>>> write('buildout.cfg',
... """
... [buildout]
... parts = zoperc
...
... [zoperc]
... recipe = zc.recipe.rhrc
... parts = zope
... dest = %(dest)s
... chkconfig = 345 90 10
... chkconfigcommand = echo
... user = root
...
... [zope]
... run-script = /opt/zope/bin/zopectl -C /etc/zope.conf
... """ % dict(dest=demo))

Then the generated script won’t su, but it will still check that root is running it:

>>> print system('bin/buildout'),
Uninstalling zoperc.
Running uninstall recipe.
--del zoperc
Installing zoperc.
--add zoperc

>>> cat(demo, 'zoperc')
#!/bin/sh
<BLANKLINE>
# the next line is for chkconfig
# chkconfig: 345 90 10
# description: please, please work
<BLANKLINE>
<BLANKLINE>
if [ $(whoami) != "root" ]; then
  echo "You must be root."
  exit 1
fi
<BLANKLINE>
case $1 in
  stop)
<BLANKLINE>
    /opt/zope/bin/zopectl -C /etc/zope.conf $* \
      </dev/null
<BLANKLINE>
    ;;
  restart)
<BLANKLINE>
    ${0} stop
    sleep 1
    ${0} start
<BLANKLINE>
    ;;
  *)
<BLANKLINE>
    /opt/zope/bin/zopectl -C /etc/zope.conf $* \
      </dev/null
<BLANKLINE>
    ;;
esac
<BLANKLINE>

A part that defines a run script can also define environment-variable settings to be used by the rc script by supplying an env option:

>>> write('buildout.cfg',
... """
... [buildout]
... parts = zoperc
...
... [zoperc]
... recipe = zc.recipe.rhrc
... parts = zope
... dest = %(dest)s
... chkconfig = 345 90 10
... chkconfigcommand = echo
... user = zope
...
... [zope]
... run-script = /opt/zope/bin/zopectl -C /etc/zope.conf
... env = LD_LIBRARY_PATH=/opt/foolib
... """ % dict(dest=demo))

>>> print system('bin/buildout'),
Uninstalling zoperc.
Running uninstall recipe.
--del zoperc
Installing zoperc.
--add zoperc

>>> cat(demo, 'zoperc')
#!/bin/sh
<BLANKLINE>
# the next line is for chkconfig
# chkconfig: 345 90 10
# description: please, please work
<BLANKLINE>
<BLANKLINE>
if [ $(whoami) != "root" ]; then
  echo "You must be root."
  exit 1
fi
<BLANKLINE>
case $1 in
  stop)
<BLANKLINE>
    LD_LIBRARY_PATH=/opt/foolib \
      su zope -c \
      "/opt/zope/bin/zopectl -C /etc/zope.conf $*" \
      </dev/null
<BLANKLINE>
    ;;
  restart)
<BLANKLINE>
    ${0} stop
    sleep 1
    ${0} start
<BLANKLINE>
    ;;
  *)
<BLANKLINE>
    LD_LIBRARY_PATH=/opt/foolib \
      su zope -c \
      "/opt/zope/bin/zopectl -C /etc/zope.conf $*" \
      </dev/null
<BLANKLINE>
    ;;
esac
<BLANKLINE>

Working with existing control scripts

In the example above, we generated a script based on a command line. If we have a part that creates a control script on it’s own, then it can ommit the run-script option and it’s already created run script will be used. Let’s create a run script ourselves:

>>> write(demo, 'zope', '/opt/zope/bin/zopectl -C /etc/zope.conf $*')

Now we can remove the run-script option from the Zope section:

>>> write('buildout.cfg',
... """
... [buildout]
... parts = zoperc
...
... [zoperc]
... recipe = zc.recipe.rhrc
... parts = zope
... dest = %(dest)s
... chkconfig = 345 90 10
... chkconfigcommand = echo
... user = zope
...
... [zope]
... env = LD_LIBRARY_PATH=/opt/foolib
... """ % dict(dest=demo))

>>> print system('bin/buildout'),
Uninstalling zoperc.
Running uninstall recipe.
--del zoperc
Installing zoperc.
--add zoperc

>>> cat(demo, 'zoperc')
#!/bin/sh
<BLANKLINE>
# the next line is for chkconfig
# chkconfig: 345 90 10
# description: please, please work
<BLANKLINE>
<BLANKLINE>
if [ $(whoami) != "root" ]; then
  echo "You must be root."
  exit 1
fi
<BLANKLINE>
case $1 in
  stop)
<BLANKLINE>
    echo zope:
/demo/zope "$@" \
      </dev/null
<BLANKLINE>
    ;;
  restart)
<BLANKLINE>
    ${0} stop
    sleep 1
    ${0} start
<BLANKLINE>
    ;;
  *)
<BLANKLINE>
    echo zope:
/demo/zope "$@" \
      </dev/null
<BLANKLINE>
    ;;
esac
<BLANKLINE>

Here we just invoke the existing script. Note that don’t pay any reflect the env or user options in the script. When an existing script is used, it is assumed to be complete.

>>> import os
>>> os.remove(join(demo, 'zope'))

Multiple processes

Sometimes, you need to start multiple processes. You can specify multiple parts. For example, suppose we wanted to start 2 Zope instances:

>>> write('buildout.cfg',
... """
... [buildout]
... parts = zoperc
...
... [zoperc]
... recipe = zc.recipe.rhrc
... parts = instance1 instance2
... dest = %(dest)s
... chkconfig = 345 90 10
... chkconfigcommand = echo
... user = zope
...
... [instance1]
... run-script = /opt/zope/bin/zopectl -C /etc/instance1.conf
... env = LD_LIBRARY_PATH=/opt/foolib
...
... [instance2]
... """ % dict(dest=demo))

>>> write(demo, 'instance2', '')

Note that for instance 2, we are arranging for the script to pre-exist.

>>> print system('bin/buildout'),
Uninstalling zoperc.
Running uninstall recipe.
--del zoperc
Installing zoperc.
--add zoperc

>>> cat(demo, 'zoperc')
#!/bin/sh
<BLANKLINE>
# the next line is for chkconfig
# chkconfig: 345 90 10
# description: please, please work
<BLANKLINE>
<BLANKLINE>
if [ $(whoami) != "root" ]; then
  echo "You must be root."
  exit 1
fi
<BLANKLINE>
case $1 in
  stop)
<BLANKLINE>
    echo instance2:
/demo/instance2 "$@" \
      </dev/null
<BLANKLINE>
    LD_LIBRARY_PATH=/opt/foolib \
      su zope -c \
      "/opt/zope/bin/zopectl -C /etc/instance1.conf $*" \
      </dev/null
<BLANKLINE>
    ;;
  restart)
<BLANKLINE>
    ${0} stop
    sleep 1
    ${0} start
<BLANKLINE>
    ;;
  *)
<BLANKLINE>
    LD_LIBRARY_PATH=/opt/foolib \
      su zope -c \
      "/opt/zope/bin/zopectl -C /etc/instance1.conf $*" \
      </dev/null
<BLANKLINE>
    echo instance2:
/demo/instance2 "$@" \
      </dev/null
<BLANKLINE>
    ;;
esac
<BLANKLINE>

Now the rc script starts both instances. Note that it stops them in reverese order. This isn’t so important in a case like this, but would be more important if a later script depended on an earlier one.

In addition to the zoperc script, we got scripts for the instance with the run-script option:

>>> ls(demo)
-  instance2
-  zoperc
-  zoperc-instance1

>>> cat(demo, 'zoperc-instance1')
#!/bin/sh
<BLANKLINE>
# This script is for adminstrator convenience.  It should
# NOT be installed as a system startup script!
<BLANKLINE>
<BLANKLINE>
if [ $(whoami) != "root" ]; then
  echo "You must be root."
  exit 1
fi
<BLANKLINE>
case $1 in
  stop)
<BLANKLINE>
    LD_LIBRARY_PATH=/opt/foolib \
      su zope -c \
      "/opt/zope/bin/zopectl -C /etc/instance1.conf $*"
<BLANKLINE>
    ;;
  restart)
<BLANKLINE>
    ${0} stop
    sleep 1
    ${0} start
<BLANKLINE>
    ;;
  *)
<BLANKLINE>
    LD_LIBRARY_PATH=/opt/foolib \
      su zope -c \
      "/opt/zope/bin/zopectl -C /etc/instance1.conf $*"
<BLANKLINE>
    ;;
esac
<BLANKLINE>

The individual scripts don’t have chkconfig information.

Independent processes

Normally, processes are assumed to be dependent and are started in order, stopped in referese order, and, on restart, are all stopped and then all started.

If the independent-processes option is used, then the generated master run script will treat the processes as independent and restart processed individually. With lots of independent processes, this can reduce the amount of time individual processes are down.

>>> write('buildout.cfg',
... """
... [buildout]
... parts = zoperc
...
... [zoperc]
... recipe = zc.recipe.rhrc
... parts = instance1 instance2
... dest = %(dest)s
... chkconfig = 345 90 10
... chkconfigcommand = echo
... user = zope
... independent-processes = true
...
... [instance1]
... run-script = /opt/zope/bin/zopectl -C /etc/instance1.conf
... env = LD_LIBRARY_PATH=/opt/foolib
...
... [instance2]
... """ % dict(dest=demo))

>>> print system('bin/buildout'),
Uninstalling zoperc.
Running uninstall recipe.
--del zoperc
Installing zoperc.
--add zoperc

>>> cat(demo, 'zoperc')
#!/bin/sh
<BLANKLINE>
# the next line is for chkconfig
# chkconfig: 345 90 10
# description: please, please work
<BLANKLINE>
<BLANKLINE>
if [ $(whoami) != "root" ]; then
  echo "You must be root."
  exit 1
fi
<BLANKLINE>
    LD_LIBRARY_PATH=/opt/foolib \
      su zope -c \
      "/opt/zope/bin/zopectl -C /etc/instance1.conf $*" \
      </dev/null
<BLANKLINE>
    echo instance2:
/demo/instance2 "$@" \
      </dev/null

Deployments

The zc.recipe.rhrc recipe is designed to work with the zc.recipe.deployment recipe. You can specify the name of a deployment section. If a deployment section is specified then:

• the deployment name will be used for the rc scripts

• the user from the deployment section will be used if a user isn’t specified in the rc script’s own section.

• the rc-directory option from the deployment will be used if destination isn’t specified.

  >>> write('buildout.cfg',
  ... """
  ... [buildout]
  ... parts = zoperc
  ...
  ... [deployment]
  ... name = acme
  ... user = acme
  ... rc-directory = %(dest)s
  ...
  ... [zoperc]
  ... recipe = zc.recipe.rhrc
  ... parts = instance1 instance2
  ... chkconfig = 345 90 10
  ... chkconfigcommand = echo
  ... deployment = deployment
  ...
  ... [instance1]
  ... run-script = /opt/zope/bin/zopectl -C /etc/instance1.conf
  ... env = LD_LIBRARY_PATH=/opt/foolib
  ...
  ... [instance2]
  ... """ % dict(dest=demo))
  

If a deployment is used, then any existing scripts must be prefixed with the deployment name. We’ll rename the instance2 script to reflect that:

>>> os.rename(join(demo, 'instance2'), join(demo, 'acme-instance2'))

>>> print system('bin/buildout'),
Uninstalling zoperc.
Running uninstall recipe.
--del zoperc
Installing zoperc.
--add acme

>>> ls(demo)
-  acme
-  acme-instance1
-  acme-instance2

>>> cat(demo, 'acme')
#!/bin/sh
<BLANKLINE>
# the next line is for chkconfig
# chkconfig: 345 90 10
# description: please, please work
<BLANKLINE>
<BLANKLINE>
if [ $(whoami) != "root" ]; then
  echo "You must be root."
  exit 1
fi
<BLANKLINE>
case $1 in
  stop)
<BLANKLINE>
    echo acme-instance2:
/demo/acme-instance2 "$@" \
      </dev/null
<BLANKLINE>
    LD_LIBRARY_PATH=/opt/foolib \
      su acme -c \
      "/opt/zope/bin/zopectl -C /etc/instance1.conf $*" \
      </dev/null
<BLANKLINE>
    ;;
  restart)
<BLANKLINE>
    ${0} stop
    sleep 1
    ${0} start
<BLANKLINE>
    ;;
  *)
<BLANKLINE>
    LD_LIBRARY_PATH=/opt/foolib \
      su acme -c \
      "/opt/zope/bin/zopectl -C /etc/instance1.conf $*" \
      </dev/null
<BLANKLINE>
    echo acme-instance2:
/demo/acme-instance2 "$@" \
      </dev/null
<BLANKLINE>
    ;;
esac
<BLANKLINE>

Edge case, when we remove the part, we uninstall acme:

>>> write('buildout.cfg',
... """
... [buildout]
... parts =
... """)
>>> print system('bin/buildout'),
Uninstalling zoperc.
Running uninstall recipe.
--del acme

Process Management

Normally, the recipe doesn’t start and stop processes. If we want it to, we can use the process-management option with a ‘true’ value.

>>> write('buildout.cfg',
... """
... [buildout]
... parts = zoperc
...
... [zoperc]
... recipe = zc.recipe.rhrc
... parts = zope
... dest = %(dest)s
... process-management = true
...
... [zope]
... run-script = echo zope
... """ % dict(dest=demo))

When the part is installed, the process is started:

>>> print system('bin/buildout'),
Installing zoperc.
zope start

It also gets started when the part updates. This is just to make sure it is running.

>>> print system('bin/buildout'),
Updating zoperc.
zope start

If we update the part, then when the part is uninstalled and reinstalled, the process will be stopped and started. We’ll often force this adding a digest option that exists solely to force a reinstall, typically because something else in the buildout has changed.

>>> write('buildout.cfg',
... """
... [buildout]
... parts = zoperc
...
... [zoperc]
... recipe = zc.recipe.rhrc
... parts = zope
... dest = %(dest)s
... process-management = true
... digest = 1
...
... [zope]
... run-script = echo zope
... """ % dict(dest=demo))

>>> print system('bin/buildout'),
Uninstalling zoperc.
Running uninstall recipe.
zope stop
Installing zoperc.
zope start

>>> print system('bin/buildout buildout:parts='),
Uninstalling zoperc.
Running uninstall recipe.
zope stop

Regression Tests

Exception formatting bug

If we do not provide a runscript, we get an exception (bug was: improperly formatted exception string, contained literal ‘%s’):

>>> write('buildout.cfg',
... """
... [buildout]
... parts = zoperc
...
... [zoperc]
... recipe = zc.recipe.rhrc
... parts = zope
... dest = %(dest)s
...
... [zope]
... """ % dict(dest=demo))
>>> print system('bin/buildout'),
Installing zoperc.
zc.recipe.rhrc: Part zope doesn't define run-script and /demo/zope doesn't exist.
While:
  Installing zoperc.
Error: No script for zope

Download