Dear community, I have got an interesting behaviour in HA Weaviate cluster, here is what we have done:
- we deleted big number of objects
- after that, we do a search and we still see ~ 1.7K objects, which we have deleted
- the number of found objects returned from the search is different (+/- 100 objects) - it looks like it depends on which cluster node the request ends
I’ve gone trough the CAP implementation consideration in Weaviate, but I doubt it is related to eventual consistency as after 24h we still observe the behaviour above.
PS: We use Weaviate Cloud service