Update AlgoliaUnreachableHostException to OperationTimeout

[dylancaponi]

• Algolia Client Version: 2.5.0 (I see that 3.0 was released 3 days ago, not sure if that resolves this)
• Language Version: Python 3.9.7

Description

Client throws algoliasearch.exceptions.AlgoliaUnreachableHostException: Unreachable hosts during replace_all_objects operation. The issue was not that hosts were unreachable but that the operation was timing out. I fixed the problem by setting readTimeout to 100:

client = SearchClient.create(algolia_application_id, algolia_api_key)
index = client.init_index(algolia_index)
request_options = {
    "readTimeout": 100
}
index.replace_all_objects(records, request_options)

A timeout should produce a timeout error. Unreachable is misleading.

Steps To Reproduce

See description. Use replace_all_objects on a sufficiently big index. Mine was 6393 rows and 10+ columns some with significant text blocks.

[shortcuts]

Hey, thanks for the issue! We throw an unreachable host error because after a timeout error, we will try an other host (Algolia's API offers multiple), then if there's no host left we end up in an AlgoliaUnreachableHostException

I can however confirm that it can be misleading when doing large enough operations

[dylancaponi]

I'm now seeing this error regardless how big I set the readTimeout. I've triple checked App ID and API Key.

[dylancaponi]

Oh, you also get this message if the index does not exist. I would expect index does not Exist rather than an unreachable hosts error...