问题
I am trying to understand what consistency guarantees Azure Search provides.
Scenario: Let's say I'm posting documents sequentially to Azure Search REST API:
- doc1
- doc2
- doc3
Questions
- I issue a query and doc3 is returned. Does it imply that doc1 and doc2 are also present in the index? E.g., should I be concerned about data replication not being completed yet, etc.?
- If POST request returns, is it 100% guaranteed that the posted document is available to search, hence will immediately be found by any client querying the index?
- Does any of the behavior related to consistency depend on the tier used?
回答1:
1) Azure Search guarantees that eventually all documents show up, but not that readers will see them show up in order (no monotonic reads).
2) No. If you see HTTP status 200 from the indexing request, it means the data is durable (we won’t lose it) but not that it’s searchable. It may take a few seconds to be searchable depending the load.
3) All tiers expose an eventual consistency model. There might be minor differences in typical timings/sequence among tiers due to differences in how data is partitioned, but these are differences that you shouldn’t take a dependency on since they are variable.
来源：https://stackoverflow.com/questions/44590456/azure-search-data-consistency-when-posting-to-index