Add logging at main API endpoints (bulk ingest, msearch) when 500 is returned

[autata]

Summary

Add 2 more places to log errors where astra returns a 500. Without this, nothing is logged in these cases when astra encounters an internal server error. I picked error level because other places where a 500 is returned, those are also logged as errors. These may not be as severe and more implementation choices. I'm curious what level the reviewers would prefer here.

In the case of request must contain only 1 unique index, debug may be a more suitable log level as there could be a lot of these and error logs could be spammy.

For multisearch query, an example where this is triggered would be shown in the PR: #780. (Null pointer exception due to json deserialization logic) Potentially, this could be better suited as a warning log?

Requirements

• I've read and understood the Contributing Guidelines and have done my best effort to follow them.
• I've read and agree to the Code of Conduct.

[bryanlb]

Ah, this is a good add.

[bryanlb]

I believe we should already be logging these, as we've enabled Armeria request logging here:

astra/astra/src/main/java/com/slack/astra/server/ArmeriaService.java

Lines 72 to 84 in 2d7c87f

	private void initializeLogging() {
	serverBuilder.decorator(
	LoggingService.builder()
	// Not logging any successful response, say prom scraping /metrics every 30 seconds at
	// INFO
	.successfulResponseLogLevel(LogLevel.DEBUG)
	.failureResponseLogLevel(LogLevel.ERROR)
	// Remove the content to prevent blowing up the logs
	.responseContentSanitizer((ctx, content) -> "truncated")
	// Remove all headers to be sure we aren't leaking any auth/cookie info
	.requestHeadersSanitizer((ctx, headers) -> DefaultHttpHeaders.EMPTY_HEADERS)
	.newDecorator());
	}

This should automatically log all successful and failed requests at the log levels indicated. This does make use of responseContentSanitizer which has since been deprecated in Armeria.

I'm not sure if this currently logs out the stack trace, but instead of adding try/catches to the individual endpoints I think that looking into the LoggingService decorator would be a better approach.

[github-actions]

This PR is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 3 days.