Review Request 723: Rewrite the registration script in preparation for further improvements

[Stephen Gallagher]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://reviewboard-openlmi.rhcloud.com/r/723/
-----------------------------------------------------------

Review request for OpenLMI Developers.


Repository: openlmi-providers


Description
-------

Here's a new registration script. It's a merge of the existing
openlmi-mof-register and openlmi-register pegasus and should work as drop-in replacement
for the openlmi-mof-register. The only slight difference is that the Pegasus interop
namespace is not hard coded into the Pegasus registration calls but now there's an
attempt to detect the right one.  And the whole thing is now written in (sort of) Python.

This is only the first step in the registration rewrite: Python should make it easier for
manipulating a "side database" that would keep the registration data
independently of the CIMOMs and eventually allow for re-registration when a new CIMOM is
installed or upgraded.  This will also make possible for the providers package not to
depend on a CIMOM.  I will prepare some design sketches of the registration process I
envision to gather feedback and ideas: doing this correctly and robustly enough is going
to be rather difficult.


Diffs
-----

  openlmi-mof-register.py PRE-CREATION 

Diff: http://reviewboard-openlmi.rhcloud.com/r/723/diff/


Testing
-------


Thanks,

Tomas Smetana

[Stephen Gallagher]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://reviewboard-openlmi.rhcloud.com/r/723/#review917
-----------------------------------------------------------



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment487>

    I think that this shouldn't be here.



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment488>

    I think that this shouldn't be here.



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment489>

    At this point, I think it would be a good idea to exit() with an error message (maybe
a broken installation?). Unless Pegasus can have no "root/interop" or
"root/PG_InterOp" and still work properly.


- Jan Synacek


On Aug. 15, 2013, 3:50 p.m., Tomas Smetana wrote:
>  
>  -----------------------------------------------------------
>  This is an automatically generated e-mail. To reply, visit:
>  http://reviewboard-openlmi.rhcloud.com/r/723/
>  -----------------------------------------------------------
>  
>  (Updated Aug. 15, 2013, 3:50 p.m.)
>  
>  
>  Review request for OpenLMI Developers.
>  
>  
>  Repository: openlmi-providers
>  
>  
>  Description
>  -------
>  
>  Here's a new registration script. It's a merge of the existing
> openlmi-mof-register and openlmi-register pegasus and should work as drop-in replacement
> for the openlmi-mof-register. The only slight difference is that the Pegasus interop
> namespace is not hard coded into the Pegasus registration calls but now there's an
> attempt to detect the right one.  And the whole thing is now written in (sort of) Python.
>  
>  This is only the first step in the registration rewrite: Python should make it easier for
> manipulating a "side database" that would keep the registration data
> independently of the CIMOMs and eventually allow for re-registration when a new CIMOM is
> installed or upgraded.  This will also make possible for the providers package not to
> depend on a CIMOM.  I will prepare some design sketches of the registration process I
> envision to gather feedback and ideas: doing this correctly and robustly enough is going
> to be rather difficult.
>  
>  
>  Diffs
>  -----
>  
>    openlmi-mof-register.py PRE-CREATION 
>  
>  Diff: http://reviewboard-openlmi.rhcloud.com/r/723/diff/
>  
>  
>  Testing
>  -------
>  
>  
>  Thanks,
>  
>  Tomas Smetana
>  
>

[Stephen Gallagher]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://reviewboard-openlmi.rhcloud.com/r/723/#review918
-----------------------------------------------------------

Ship it!



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment490>

    According to PEP8: "version == None" is wrong, use "version is
None".
    
    http://www.python.org/dev/peps/pep-0008/#programming-recommendations


Looks good. Feel free to ship it now or add the database first.

- Radek Novacek


On Aug. 15, 2013, 3:50 p.m., Tomas Smetana wrote:
>  
>  -----------------------------------------------------------
>  This is an automatically generated e-mail. To reply, visit:
>  http://reviewboard-openlmi.rhcloud.com/r/723/
>  -----------------------------------------------------------
>  
>  (Updated Aug. 15, 2013, 3:50 p.m.)
>  
>  
>  Review request for OpenLMI Developers.
>  
>  
>  Repository: openlmi-providers
>  
>  
>  Description
>  -------
>  
>  Here's a new registration script. It's a merge of the existing
> openlmi-mof-register and openlmi-register pegasus and should work as drop-in replacement
> for the openlmi-mof-register. The only slight difference is that the Pegasus interop
> namespace is not hard coded into the Pegasus registration calls but now there's an
> attempt to detect the right one.  And the whole thing is now written in (sort of) Python.
>  
>  This is only the first step in the registration rewrite: Python should make it easier for
> manipulating a "side database" that would keep the registration data
> independently of the CIMOMs and eventually allow for re-registration when a new CIMOM is
> installed or upgraded.  This will also make possible for the providers package not to
> depend on a CIMOM.  I will prepare some design sketches of the registration process I
> envision to gather feedback and ideas: doing this correctly and robustly enough is going
> to be rather difficult.
>  
>  
>  Diffs
>  -----
>  
>    openlmi-mof-register.py PRE-CREATION 
>  
>  Diff: http://reviewboard-openlmi.rhcloud.com/r/723/diff/
>  
>  
>  Testing
>  -------
>  
>  
>  Thanks,
>  
>  Tomas Smetana
>  
>

[Stephen Gallagher]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://reviewboard-openlmi.rhcloud.com/r/723/#review926
-----------------------------------------------------------



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment510>

    This could be turned to docopt parseable usage string. It's pity we don't want
to depend on it :(.
    
    Anyway it would be nice to use some smart argument parser, that would do the hard work
of checking for us. For example argparse. But this could be done later.



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment506>

    I'd probably move the body of conditions to separate functions
(unregister_sfcbd(), unregister_pegasus()).



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment509>

    Since ``devnull`` is handy on multiple places like this, I think it would make sense
to make it a global variable.
    
    Or make an override for ``subprocess.call()`` like this:
    
    def call(*args, **kwargs):
      with open("/dev/null", "w") as devnull:
        kwargs['stdout'] = devnull
        kwargs['stderr'] = devnull
        return subprocess.call(*args, **kwargs)
    
    usage:
      ret = call(["/usr/sbin/cimserver", "--status"])



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment508>

    Please use:
      with open("regfile", "w") as regfile:
         # guarded code with ``regfile`` available
      # no explicit ``close()`` call needed
    
    It's more readable and robust.
    It applies to other open() calls as well.


- Michal Minar


On Aug. 15, 2013, 1:50 p.m., Tomas Smetana wrote:
>  
>  -----------------------------------------------------------
>  This is an automatically generated e-mail. To reply, visit:
>  http://reviewboard-openlmi.rhcloud.com/r/723/
>  -----------------------------------------------------------
>  
>  (Updated Aug. 15, 2013, 1:50 p.m.)
>  
>  
>  Review request for OpenLMI Developers.
>  
>  
>  Repository: openlmi-providers
>  
>  
>  Description
>  -------
>  
>  Here's a new registration script. It's a merge of the existing
> openlmi-mof-register and openlmi-register pegasus and should work as drop-in replacement
> for the openlmi-mof-register. The only slight difference is that the Pegasus interop
> namespace is not hard coded into the Pegasus registration calls but now there's an
> attempt to detect the right one.  And the whole thing is now written in (sort of) Python.
>  
>  This is only the first step in the registration rewrite: Python should make it easier for
> manipulating a "side database" that would keep the registration data
> independently of the CIMOMs and eventually allow for re-registration when a new CIMOM is
> installed or upgraded.  This will also make possible for the providers package not to
> depend on a CIMOM.  I will prepare some design sketches of the registration process I
> envision to gather feedback and ideas: doing this correctly and robustly enough is going
> to be rather difficult.
>  
>  
>  Diffs
>  -----
>  
>    openlmi-mof-register.py PRE-CREATION 
>  
>  Diff: http://reviewboard-openlmi.rhcloud.com/r/723/diff/
>  
>  
>  Testing
>  -------
>  
>  
>  Thanks,
>  
>  Tomas Smetana
>  
>

[Stephen Gallagher]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://reviewboard-openlmi.rhcloud.com/r/723/#review927
-----------------------------------------------------------


Thanks everyone for the comments.


openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment518>

    It depends on the "depth" of the dependency: This is an utility script and I
would like to avoid it to drag in too many dependencies of its own. ([Disclaimer] I have
no idea what docopt is: I tend to write a C code no matter what language I use...)



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment511>

    There is a commented-out subprocess-call that makes use of the devnull. It's
mimicking the behaviour of the original openlmi-mof-register script bu I found it useful
to see the stderr of the called program as well. You're right that this part will need
some ipmrovements: Michal has suggested a way to fix all the /dev/null redirected
subprocess invocations.



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment512>

    dtto



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment513>

    Right. I will fix this.



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment514>

    I don't disagree. However for now I would prefer to have the script easily
comparable to the original one before I'm reasonably sure I didn't introduce any
new bugs.



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment515>

    Looks nice. Thanks. Will fix that.



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment517>

    Will do.



openlmi-mof-register.py
<http://reviewboard-openlmi.rhcloud.com/r/723/#comment516>

    Fixed. Will post a new version later.


- Tomas Smetana


On Aug. 15, 2013, 3:50 p.m., Tomas Smetana wrote:
>  
>  -----------------------------------------------------------
>  This is an automatically generated e-mail. To reply, visit:
>  http://reviewboard-openlmi.rhcloud.com/r/723/
>  -----------------------------------------------------------
>  
>  (Updated Aug. 15, 2013, 3:50 p.m.)
>  
>  
>  Review request for OpenLMI Developers.
>  
>  
>  Repository: openlmi-providers
>  
>  
>  Description
>  -------
>  
>  Here's a new registration script. It's a merge of the existing
> openlmi-mof-register and openlmi-register pegasus and should work as drop-in replacement
> for the openlmi-mof-register. The only slight difference is that the Pegasus interop
> namespace is not hard coded into the Pegasus registration calls but now there's an
> attempt to detect the right one.  And the whole thing is now written in (sort of) Python.
>  
>  This is only the first step in the registration rewrite: Python should make it easier for
> manipulating a "side database" that would keep the registration data
> independently of the CIMOMs and eventually allow for re-registration when a new CIMOM is
> installed or upgraded.  This will also make possible for the providers package not to
> depend on a CIMOM.  I will prepare some design sketches of the registration process I
> envision to gather feedback and ideas: doing this correctly and robustly enough is going
> to be rather difficult.
>  
>  
>  Diffs
>  -----
>  
>    openlmi-mof-register.py PRE-CREATION 
>  
>  Diff: http://reviewboard-openlmi.rhcloud.com/r/723/diff/
>  
>  
>  Testing
>  -------
>  
>  
>  Thanks,
>  
>  Tomas Smetana
>  
>

[Stephen Gallagher]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://reviewboard-openlmi.rhcloud.com/r/723/
-----------------------------------------------------------

(Updated Oct. 3, 2013, 10:42 a.m.)


Status
------

This change has been discarded.


Review request for OpenLMI Developers.


Repository: openlmi-providers


Description
-------

Here's a new registration script. It's a merge of the existing
openlmi-mof-register and openlmi-register pegasus and should work as drop-in replacement
for the openlmi-mof-register. The only slight difference is that the Pegasus interop
namespace is not hard coded into the Pegasus registration calls but now there's an
attempt to detect the right one.  And the whole thing is now written in (sort of) Python.

This is only the first step in the registration rewrite: Python should make it easier for
manipulating a "side database" that would keep the registration data
independently of the CIMOMs and eventually allow for re-registration when a new CIMOM is
installed or upgraded.  This will also make possible for the providers package not to
depend on a CIMOM.  I will prepare some design sketches of the registration process I
envision to gather feedback and ideas: doing this correctly and robustly enough is going
to be rather difficult.


Diffs
-----

  openlmi-mof-register.py PRE-CREATION 

Diff: http://reviewboard-openlmi.rhcloud.com/r/723/diff/


Testing
-------


Thanks,

Tomas Smetana