Details
-
Improvement
-
Status: Closed
-
Top
-
Resolution: Fixed
-
None
-
None
-
None
Description
For each nodetype that gets reloaded like in the hipposys namespace the complete nodetype definition gets saved in the repository as a binary xml file. This makes starting up very slow. Also the other nodes in the cluster have to parse and update there nodetype information. Thirdly it's error prone, we shouldn't write (critical) data if nothing has changed. We have seen bugs in the past that an empty nodetype def got written to the database.
In my opinion the (system) namespaces should only be reload if they have changed. Also the namespace should be reloaded per namespace and not per nodetype.
Attachments
Issue Links
- relates to
-
REPO-262 Investigate clustered startup node type redefinition problems
- Closed
-
REPO-144 Error during startup with clustered repositories
- Closed
-
REPO-273 Restart of a clustered node causes data corruption when node types are reregistered
- Closed
-
REPO-1699 Ensure nodetype definitions are only (re)loaded when the cnd resource file has been changed
- Closed