Open ylwu-amzn opened 2 months ago
Change to implement SearchResponseProcessor.processResponseAsync
method should be good,
@austintlee , we need to prioritize this issue. Do you have enough bandwidth ?
cc @mashah @jonfritz
@austintlee Find some other places
CreateConversationResponse response = client
.execute(CreateConversationAction.INSTANCE, new CreateConversationRequest(name))
.actionGet(DEFAULT_TIMEOUT_IN_MILLIS);
CreateInteractionResponse res = client
.execute(
CreateInteractionAction.INSTANCE,
new CreateInteractionRequest(conversationId, input, promptTemplate, response, origin, additionalInfo)
)
.actionGet(DEFAULT_TIMEOUT_IN_MILLIS);
GetInteractionsResponse response = client
.execute(GetInteractionsAction.INSTANCE, new GetInteractionsRequest(conversationId, maxResults, from))
.actionGet(DEFAULT_TIMEOUT_IN_MILLIS);
What is the bug? Test on a multi-node cluster and see timeout exception randomly frequently
How can one reproduce the bug? Steps to reproduce the behavior:
What is the expected behavior? From error log, it's caused by https://github.com/opensearch-project/ml-commons/blob/2.11/search-processors/src/main/java/org/opensearch/searchpipelines/questionanswering/generative/client/ConversationalMemoryClient.java#L98
Should not use
actionGet
which will block thread. Suggest change to use action listenerWhat is your host/environment?
Do you have any screenshots? No
Do you have any additional context? Error log