Orphan indexes after failed reindex should be cleaned up
mbertrand opened this issue · 0 comments
mbertrand commented
Steps to Reproduce
Run recreate_index --all
with a low timeout that will cause it to fail; or time the run so that it tries to delete the old backing index as the last step when a snapshot is running (typically 20 past the hour on RC?).
Expected Behavior
The new backing index is deleted if reindex fails during initial indexing steps.
Deleting the old backing index (last step of a successful reindex) is retried multiple times if anything goes wrong.
Actual Behavior
Orphaned indexes are left behind.