preventing ldap principal to be deleted

ipa-server-install --uninstall command is calling server-del to
delete replica. This scenario does not work since server-del
is also deleting all principals from and ldap breaking ldap
replication. As a result, only part of deletions are propagated
to the other replicals leaving a lot of orphaned data there.

https://pagure.io/freeipa/issue/7371

This patch won't fully fix the issue with left-over data
but more data is cleaned up and only ldap principal is left
thus ending in a better state.
Issue will be fully fixed only when topology plugin is patched
as well. The following pagure ticket is created to track
topology plugin change:
https://pagure.io/freeipa/issue/7359

Reviewed-By: Alexander Bokovoy <abokovoy@redhat.com>

ipaserver/plugins/server.py

@@ -659,10 +659,26 @@ class server_del(LDAPDelete):
         delete server kerberos key and all its svc principals
         """
         try:
+            # do not delete ldap principal if server-del command
+            # has been called on a machine which is being deleted
+            # since this will break replication.
+            # ldap principal to be cleaned later by topology plugin
+            # necessary changes to a topology plugin are tracked
+            # under https://pagure.io/freeipa/issue/7359
+            if master == self.api.env.host:
+                filter = (
+                    '(&(krbprincipalname=*/{}@{})'
+                    '(!(krbprincipalname=ldap/*)))'
+                    .format(master, self.api.env.realm)
+                )
+            else:
+                filter = '(krbprincipalname=*/{}@{})'.format(
+                    master, self.api.env.realm
+                )
+
             entries = ldap.get_entries(
-                self.api.env.basedn, ldap.SCOPE_SUBTREE,
-                filter='(krbprincipalname=*/{}@{})'.format(
-                    master, self.api.env.realm))
+                self.api.env.basedn, ldap.SCOPE_SUBTREE, filter=filter
+            )
 
             if entries:
                 entries.sort(key=lambda x: len(x.dn), reverse=True)