arina-ielchiieva commented on issue #2010: DRILL-7620: Fix plugin mutability issues URL: https://github.com/apache/drill/pull/2010#issuecomment-597604478 @paul-rogers > if you run the UI scenario from your earlier comment, you should see that the plugin is rejected if invalid and enabled. The config is stored, but not instantiated, if invalid and disabled. I like this behavior, +1 for this approach. > If a plugin ever fails, should it be disabled? How would this work for a temporary outage? The user would have to manually re-enable the plugin when it becomes valid again. I think in this case we should NOT disable the plugin. Once DB will be back online, assume we have rdbms plugin, user will be able to query using this plugin and no additional actions will be required. The problem in disabling is that only admin has permission to enable / disable plugins, so on Production environments real users would first need to know when DB is back online and then ask admins to enable the plugin, which maybe not really convenient. > On a completely unrelated topic, there is code in the original version that forces all names of stored plugins to lower case (removing any mixed-case versions and replacing them), but no code enforces the lower-case rule in the in-memory cache. (I have tests that specifically test this case.) However, names are forced to lower case in storage. This seems more on the "bug" side of the ledger than the "feature" side. Should we allow mixed case everywhere or nowhere? Yet another thing to fix in the next PR. We should enforce lower-case plugin names everywhere.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
