[WT-9307] Recreating a dropped tiered table, that was never flushed, panics - MongoDB Jira

XML

Word

Printable

JSON

If a small change is made to test_tiered07.py to move a call to checkpoint, the test drops core:

diff --git a/test/suite/test_tiered07.py b/test/suite/test_tiered07.py

index a7636121c..02681d409 100644

--- a/test/suite/test_tiered07.py

+++ b/test/suite/test_tiered07.py

@@ -94,9 +94,9 @@ class test_tiered07(wttest.WiredTigerTestCase, TieredConfigMixin):

         c = self.session.open_cursor(self.localuri)

         c["0"] = "0"

         c.close()

-        self.session.checkpoint()

         self.pr('After data, call flush_tier')

         self.session.flush_tier(None)

+        self.session.checkpoint()

         # Drop table.

         self.pr('call drop')

Some things about this:

If the second checkpoint is just added, instead of removing the one before flush_tier the test still works.
The test is failing much later. The abort call happens in the part of the test where we try to create a new table of the same name.
The abort is happening in the tiered server thread getting an error in the copy storage operation.