11:18 a.m.
Modified with changes based on comments in #irc.
1. PKIException now takes instance and subsystem objects instead of
names.
2. Changed __repr__ for PKIInstance
3. Fixed get_tracker
4. Redefine con_dir in terms of base_dir
Tested with d9 and d10 instances. All still looks good.
On Mon, 2013-04-22 at 09:54 -0400, Ade Lee wrote:
On Fri, 2013-04-19 at 23:20 -0500, Endi Sukma Dewata wrote:
> On 4/19/2013 2:32 PM, Ade Lee wrote:
> >
> > Added servlet to return 501 for rest operations for d9 instances
> >
> > D9 instances run on tomcat6, which does not have support for the
> > autheticator and realm. We are not supporting the REST operations
> > on D9 style instances. They will need to be migrated.
> >
> > The migration framework has been modified to process d9 or d10
> > style instances, and a migration script has been added to add the new
> > servlet to existing d9 instances.
>
> Some comments (some of which have been discussed over IRC):
>
> 1. To be consistent the pki_subsystem and pki_instance classes should be
> called PKISubsystem and PKIInstance.
done
>
> 2. In PKIUpgrader's constructor the instance/subsystem objects are only
> created for validation then discarded. It might be better to move the
> validation inside the instances() or subsystems() where the objects are
> actually created and stored.
>
done - we now have a validate() method which is called by the
constructor.
> 3. There seems to be a bug in these lines:
>
> inst = pki.pki_instance(instance, instance_type)
> ...
> subs = pki.pki_instance(instance, subsystem, instance_type)
>
> The second line probably should have been calling pki.pki_subsystem()
> and the parameter should have been inst (the instance object, not the name).
>
Done. This is validation code that has been removed.
> 4. To improve clarity, the subsystem/instance name variable should be
> called 'name' inside the corresponding class, and 'subsystemName' or
> 'instanceName' outside the class.
>
Done
> 5. The PKISubsystem constructor shouldn't need to take the type
> parameter because it can be obtained from the instance.
>
Done
> 6. The message in RESTServlet could be simplified into something like this:
>
> The REST services are not available because this server is a legacy
> Dogtag 9 server. To access the REST services this server must be
> migrated into a new Dogtag 10 server.
>
Done
> 7. Instead of overriding doGet() and doPost() it's also possible to just
> override service().
>
Done
> 8. The --instance-type parameter works differently from the --instance
> and --subsystem parameters. By default pki-upgrade will upgrade all
> instances and subsystem on the system. If the --instance or --subsystem
> is specified, it will narrow the scope to the specified
> instance/subsystem only. The --instance-type, on the other hand, works
> as an additional parameter to --instance and has a default value of 10.
>
> Suppose there is a mix of Dogtag 9 and Dogtag 10 instances and some may
> have the same names (e.g. D9 pki-ca and D10 pki-ca). The parameters will
> work like this:
>
> pki-upgrade ==> upgrade all instances
>
> pki-upgrade --instance pki-ca ==> upgrade D10 pki-ca only
>
> pki-upgrade --instance pki-ca --instance-type 9
> ==> upgrade D9 pki-ca only
>
> To be consistent it should work like this:
>
> pki-upgrade --instance pki-ca ==> upgrade both pki-ca instances
>
> pki-upgrade --instance pki-ca --instance-type 10
> ==> upgrade D10 pki-ca only
>
> pki-upgrade --instance-type 9 ==> upgrade all D9 instances
>
Done
> 9. We discussed about putting the context XML in
> <instance>/conf/Catalina/localhost instead of META-INF. Is this going to
> be handled by a separate upgrade script?
>
To be handled as a separate upgrade script when we do the 10.1
customization work.
> 10. Document root could be obtained using self.doc.getroot() instead of
> self.doc.find('.').
>
Done
> 11. Element index could be obtained using self.root.index(mapping)
> instead of list(self.root).index(mapping).
>
Done
> 12. There are some trailing whitespaces.
>
Fixed