ICEBERG-13343: fix race condition in JdbcCatalog #13345
Conversation
jbonofre
left a comment
There was a problem hiding this comment.
Imho, this PR is not relevant related to previous discussions and logic used in the init method.
|
Based on @jbonofre comments I have reworked the code to not use "IF NOT EXISTS" and instead to check if the table was created after the creation fails. This case will occur during the race condition. Added Apache Derby as a test DB as SqlLite does not handle multiple connections well. Refactored some code to make it cleaner and easier to read. |
kevinjqliu
left a comment
There was a problem hiding this comment.
Thanks for adding this change to address the race condition during JDBC catalog initialization.
I think JB made a good point in the other thread,
We are suppose to create table at JDBC Catalog init, before the sink is "running".
This is a good workaround, and this PR eliminates the need for it.
I added a few nit comments. Please take a look
| } | ||
| } finally { | ||
| DriverManager.deregisterDriver(slowDriver); | ||
| } |
There was a problem hiding this comment.
nit: add a catalog operation here to verify that both catalog clients are initialized and ready
There was a problem hiding this comment.
I am not sure what you are asking for here. the future executes creates and initializes the JdbcCatalog which does not return from initialize until the initialization is complete. So the future.get at line 227 ensures that the creation and initialization has completed for both.
I don't see a Catalog method to call to verify that it is valid/working. Do you have a suggestion?
There was a problem hiding this comment.
So the test here is testing that the JdbcCatalog be can initialize with multiple concurrent connections.
My ask is to add a catalog operation (i.e. listNamespaces) on L232, after testing the initialization logic, to test that the catalog is indeed initialized correctly
| String testingDB = "jdbc:slow:derby:memory:testDb;create=true"; | ||
| new org.apache.derby.jdbc.EmbeddedDriver(); | ||
| properties.put(CatalogProperties.URI, testingDB); | ||
| SlowDriver slowDriver = new SlowDriver(testingDB); |
There was a problem hiding this comment.
Curious if we can test this without SlowDriver.
Can we simulate the same behavior by raising the parallelism?
For example, testConcurrentConnections uses 7 threads
There was a problem hiding this comment.
You can try but then the pass/fail of the test is dependent upon the system on which it runs. I feel that if you want to test a condition where a thread gets paused between two steps, make it pause.
I would be willing to make the entire wrapper classes a library so that here it would simply be creating an instance and overriding the two methods.
| } | ||
|
|
||
| @Test | ||
| public synchronized void testInitializeWithSlowConcurrentConnections() |
Co-authored-by: Fokko Driesprong <fokko@apache.org>
|
I do not have authority to merge as I do not have write access. |
|
Lgtm But only for me |
|
Thanks @Claudenw for fixing this 🙌 and thanks @jbonofre and @kevinjqliu for the review 🚀 |
* Fix for ICEBERG-13343 * removed dependency on 'IF NOT EXIST' * updated to fix checkstyle, etc. * Update core/src/main/java/org/apache/iceberg/jdbc/JdbcCatalog.java Co-authored-by: Fokko Driesprong <fokko@apache.org> --------- Co-authored-by: Fokko Driesprong <fokko@apache.org>
* Fix for ICEBERG-13343 * removed dependency on 'IF NOT EXIST' * updated to fix checkstyle, etc. * Update core/src/main/java/org/apache/iceberg/jdbc/JdbcCatalog.java Co-authored-by: Fokko Driesprong <fokko@apache.org> --------- Co-authored-by: Fokko Driesprong <fokko@apache.org>
Fix for ICEBERG-13343
Closes #13343