2.4 HA cluster broken

mist1

Hello,
I'm trying to test the configuration of the HA cluster.
After HA configuration on master and backup server, the configuration settings (VIP, rules etc) to the backup server is not synchronized. State table transfer good.
Run on master /etc/rc.filter_synchronize
Log on master

Time	Process	PID	Message
Nov 2 09:23:14	php-cgi		rc.filter_synchronize: XMLRPC reload data success with http://192.168.201.2:80/xmlrpc.php (pfsense.filter_configure).
Nov 2 09:23:11	php-cgi		rc.filter_synchronize: Beginning XMLRPC sync data to http://192.168.201.2:80/xmlrpc.php.
Nov 2 09:23:11	php-cgi		rc.filter_synchronize: XMLRPC reload data success with http://192.168.201.2:80/xmlrpc.php (pfsense.restore_config_section).
Nov 2 09:23:11	php-cgi		rc.filter_synchronize: Beginning XMLRPC sync data to http://192.168.201.2:80/xmlrpc.php.
Nov 2 09:23:11	php-cgi		rc.filter_synchronize: XMLRPC versioncheck:15.6 -- 15.6
Nov 2 09:23:11	php-cgi		rc.filter_synchronize: XMLRPC reload data success with http://192.168.201.2:80/xmlrpc.php (pfsense.host_firmware_version).
Nov 2 09:23:10	php-cgi		rc.filter_synchronize: Beginning XMLRPC sync data to http://192.168.201.2:80/xmlrpc.php.

Log on backup

Nov 2 09:23:11	php-fpm	71579	/xmlrpc.php: Disallowing CARP sync loop

PiBa

Made sure that on the backup machine the XMLRPC sync ip is left empty?

mist1

Yes, xmlrpc settings is empty on backup of course.
From my point of view maybe something is wrong in the logic at xmlrpc.php after commit
https://github.com/pfsense/pfsense/commit/f81e7cc404bc3ede065c8c6315fdb5d26b2cfc5b

public function __construct() {
		global $config;
		$this->remote_addr = $_SERVER['REMOTE_ADDR'];
		/* grab sync to ip if enabled */
		if (isset($config['hasync']['synchronizetoip']) &&
		    $config['hasync']['synchronizetoip'] == $remote_addr) {
			$this->loop_detected = true;
		}
	}

If i change this part to "force false" on a backup server xmlrpc sync work as expected

PiBa

Ah good find, broken indeed..

It sets "$this->remote_addr" but reads "$remote_addr" which would be empty.. Not exactly sure why the 'empty==empty' check allowed me to properly sync before though.. After setting the syncip on the backup and removing it again it now complains like you report.. Fix send: https://github.com/pfsense/pfsense/pull/3215

mist1

Thanks, i understood and i can check this on monday.

Ok, checked. Loop detection fixed. Wait until the PR will be commit.

jimp

I tested and merged the PR, next snaps should have it.