[Freeipa-devel] [PATCH] Don't load the LDAP schema during startup

[Jan Zeleny]

Jan Zelený <jzeleny at redhat.com> wrote:
> Jan Zelený <jzeleny at redhat.com> wrote:
> > Rob Crittenden <rcritten at redhat.com> wrote:
> > > Jan Zelený wrote:
> > > > Rob Crittenden<rcritten at redhat.com>  wrote:
> > > >> Jan Zelený wrote:
> > > >>> Loading of the schema is now performed in the first request that
> > > >>> requires it.
> > > >>> 
> > > >>> https://fedorahosted.org/freeipa/ticket/583
> > > >>> 
> > > >>> Jan
> > > >> 
> > > >> We still need to enforce that we get the schema, some low-level
> > > >> functions depend on it. Also, if the UI doesn't get its aciattrs
> > > >> (which are derived from the schema) then nothing will be editable.
> > > >> 
> > > >> I'm getting this backtrace if I force no schema by disabling
> 
> get_schema:
> > > > Ok, I'm sending new version, it should handle these exceptions better
> > > > and the operation should fail if it needs the schema and the schema
> > > > is not available for some reason.
> > > 
> > > This breaks the XML-RPC server. I fixed one problem:
> > > --- a/ipaserver/plugins/ldap2.py
> > > +++ b/ipaserver/plugins/ldap2.py
> > > 
> > > @@ -253,9 +253,10 @@ class ldap2(CrudBackend, Encoder):
> > >       def get_syntax(self, attr, value):
> > >           if not self.schema:
> > > -            self.schema = get_schema(self.ldap_uri, self.conn)
> > > -            if not self.schema:
> > > +            schema = get_schema(self.ldap_uri, self.conn)
> > > 
> > > +            if not schema:
> > >                   return None
> > > 
> > > +            object.__setattr__(self, 'schema', schema)
> > > 
> > >           obj = self.schema.get_obj(_ldap.schema.AttributeType, attr)
> > >           
> > >           if obj is not None:
> > >               return obj.syntax
> > > 
> > > But simply things like get_entry() return an InternalError now. I'm not
> > > sure where you were going by adding this.
> > > 
> > > rob
> > 
> > Ok, no problem. It's possible that I simply did a mistake thinking I can
> > do something in Python what is not really possible.
> > 
> > About that InternalError: I think raising InternalError when we cannot
> > load the schema to do the decoding is the right thing to do. Do you have
> > a better solution? I thought about returning empty result, but that
> > would mean we have to check the result in every funtction that is
> > calling them and raising InternalError there.
> 
> I'm sending updated patch. I modified the get_syntax() as you suggested and
> I slightly modified raising that InternalError - currently it isn't raised
> when results from get_entry() are not required by calling method.
> Currently I'm running some tests, preliminary results looked ok.

self-nack

I discovered some issues discovered by internal test suite, I'm working on 
them

Jan