Bugs in ElasticSearch request error reporting

[tcrsavage]

Steps to reproduce:

1. Delete a post, triggering immediate deletion request to ES (this does not go through the ES bulk request queue) - triggers false positive error PHP Warning: Error in ElasticSearch request: (0) in /usr/src/app/vendor/altis/cloud/inc/namespace.php on line 379 with no debug data. In this case should probably trigger a warning that the request is non blocking, thus response cannot be inspected for errors

2. Send a large request which triggers split_large_ep_request, this will bypass request logging due to direct use of wp_remote_request. Logging requests on a lower level should solve this: see example below

add_action( 'http_api_debug', __NAMESPACE__ . '\\enhanced_es_debug', 10, 5 );

/**
 * Do an error log of failed ES requests
 * Lower level than inbuilt altis one, as it is able to catch any request made to ES via wp_remote_{x}
 *
 * @param array|WP_Error $response
 * @param string $context
 * @param string $class
 * @param array $parsed_args
 * @param string $url
 * @return void
 */
function enhanced_es_debug( $response, $context, $class, $parsed_args, $url ) {
	if ( $context !== 'response' ) {
		return;
	}

	// phpcs:ignore WordPress.WP.AlternativeFunctions.parse_url_parse_url
	$host = parse_url( $url, PHP_URL_HOST );

	if ( ELASTICSEARCH_HOST !== $host ) {
		return;
	}

	$request_response_body = wp_remote_retrieve_body( $response );
	$request_response_code = (int) wp_remote_retrieve_response_code( $response );
	$is_valid_res = ( $request_response_code >= 200 && $request_response_code <= 299 );
	$type = $parsed_args['method'] ?? 'GET?';

	// Backup check for errors, sometimes the response is ok but the query
	// response JSON contains errors.
	$has_errors = strpos( $request_response_body, '"errors":true' ) !== false;

	if ( is_wp_error( $response ) ) {
		// phpcs:ignore WordPress.Security.EscapeOutput.OutputNotEscaped
		trigger_error( sprintf( 'RBMH Error in ElasticPress request: WP_Error %s %s (%s) %s', $type, $response->get_error_message(), $response->get_error_code(), $url ), E_USER_WARNING );
	} elseif ( ! $is_valid_res || $has_errors ) {
		// phpcs:ignore WordPress.Security.EscapeOutput.OutputNotEscaped
		trigger_error( sprintf( 'RBMH Error in ElasticPress request: %s %s (%s) %s', $type, $request_response_body, $request_response_code, $url ), E_USER_WARNING );
	}
}

[itc-michael-tran]

Hi @tcrsavage , did you solve this bug?
I get the same error while trying to save my ACF fields

[tcrsavage]

Hi Micheal The error log you are seeing there is a false positive, the request being sent is a DELETE call which isn't waiting on a response from the elasticsearch server, and because it doesn't wait for a response, the error logging code in altis assumes that no response means something went wrong If you're debugging an issue, it's probably unrelated to this error log If you want to stop this error from appearing in your logs, ideally the issue gets patched in altis, otherwise I think you could unhook (remove_action) the existing altis logging function and replace it with the one I added here. Let me know if you need help with that

…

On Wed, 7 Feb 2024, 02:25 itc-michael-tran, ***@***.***> wrote: Hi @tcrsavage <https://github.com/tcrsavage> , did you solve this bug? I get the same error while trying to save my ACF fields — Reply to this email directly, view it on GitHub <#371 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAG5SAK7HEXPKE24DJSR4GDYSLQYDAVCNFSM5M2YWLSKU5DIOJSWCZC7NNSXTN2JONZXKZKDN5WW2ZLOOQ5TCOJTGEYTGNZZGUYA> . You are receiving this because you were mentioned.Message ID: ***@***.***>