Hi, this morning we saw requests to Confluence search failing - the incident seemed to be resolved pretty quickly 
We have captured a response with body, which looked like this:
400 Bad Request
{
"statusCode" : 400,
"data" : {
"authorized" : false,
"valid" : true,
"errors" : [ ],
"successful" : false
},
"message" : "com.atlassian.confluence.api.service.exceptions.BadRequestException:
CQL was parsed but the search manager was unable to execute the search. CQL: '/*redacted :) */'.
Error message: java.lang.RuntimeException: Hystrix circuit short-circuited and is OPEN"
}
I was wondering why we get a 400 here, which means we did something wrong. As the message exposes there’s a circuit breaker in open state, so I would hope that situation gets resolved pretty quick - ideally even automatically.
So getting a 429 or 503 might enable API users to retry the requests. For a background/batch task that runs for hours it can make all the difference if it can wait a few minutes (using exponential backoff etc. - see Rate limiting) instead of failing and having to start over.
Any comments or plans to change that?