In a Mariadb cluster (4 node) with gtid mode on we have problems with causal_reads. After an insert, the next query waits until the causal_reads_timeout reaches its maximum value. We are not sure if this problem comes from Maxscale or Percona.

Steps to reproduce:

CREATE TABLE `test_lrl`.`test` (
`id` INT NOT NULL AUTO_INCREMENT,
`text` VARCHAR(45) NULL,
PRIMARY KEY (`id`));

SELECT id,text,@@HOSTNAME FROM test_lrl.test;
12:39:43 SELECT id,text,@@HOSTNAME FROM test_lrl.test LIMIT 0, 1000 3 row(s) returned *0,0094 *sec / 0,000017 sec

INSERT INTO `test_lrl`.`test` (`text`) VALUES ('Test');

SHOW SESSION VARIABLES WHERE Variable_name = 'last_gtid';
maxscalen1:> 'last_gtid', '1-1-480601'

Mariadb nodes: SHOW GLOBAL VARIABLES LIKE 'gtid_current_pos';
n1:> 'last_gtid', '1-1-480639'
n2:> 'last_gtid', '1-1-480639'
n3:> 'last_gtid', '1-1-480639'
n4:> 'last_gtid', '1-1-480639'

SELECT id,text,@@HOSTNAME FROM test_lrl.test;
12:41:10 SELECT id,text,@@HOSTNAME FROM test_lrl.test LIMIT 0, 1000 4 row(s) returned *20,014 *sec / 0,000015 sec
'13', 'Test', 'db-c01-n3'

Mariadb nodes:
n1:>
n2:> Waiting in MASTER_GTID_WAIT() (primary waiter) | SET @maxscale_secret_variable=(SELECT CASE WHEN MASTER_GTID_WAIT('1-1-480601', 20) = 0 THEN 1 ELSE (SELECT 1 FROM INFORMATION_SCHEMA.ENGINES) END)
n3:>
n4:>

Config Maxscale:

--------------------------------------------------------------------
Server | Address | Port | Connections | Status
--------------------------------------------------------------------
db-c01-n1 | 10.1.1.162 | 3306 | 2 | Slave, Synced, Running
db-c01-n2 | 10.1.1.163 | 3306 | 1 | Slave, Synced, Running
db-c01-n3 | 10.1.1.164 | 3306 | 4 | Master, Synced, Running
db-c01-n4 | 10.1.1.165 | 3306 | 1 | Slave, Synced, Running
--------------------------------------------------------------------

[galera-service]
type=service
router=readwritesplit
servers=db-c01-n1,db-c01-n2,db-c01-n3,db-c01-n4,
user=maxscale
password=******
localhost_match_wildcard_host=1
use_sql_variables_in=master
max_sescmd_history=0
disable_sescmd_history=true
filters=NamedServerFilter
max_slave_connections=1
slave_selection_criteria=LEAST_GLOBAL_CONNECTIONS
causal_reads=1
causal_reads_timeout=20

[NamedServerFilter]
type=filter
module=namedserverfilter
match01=.temp\..
target01=->master
match02=.`temp`\..
target02=->master

Galera cluster:
[mysqld]
binlog_format=ROW
innodb_autoinc_lock_mode=2
bind-address=0.0.0.0

wsrep_on=ON
wsrep_provider=/usr/lib64/galera/libgalera_smm.so

wsrep_cluster_name="mariadb-c01-test"
wsrep_cluster_address="gcomm://db-c01-n1,db-c01-n2,db-c01-n3,db-c01-n4"

wsrep_sst_method=mariabackup
wsrep_sst_auth=replication:******

wsrep_node_address="10.1.1.162"
wsrep_node_name="db-c01-n1"

1. Settings for gtid cluster ():
   wsrep_slave_threads=8
   wsrep_gtid_mode=ON
   wsrep_gtid_domain_id=1
   log_slave_updates=ON
   log_bin=db-c01
   expire_logs_days=1
   session_track_system_variables=autocommit,character_set_client,character_set_connection,character_set_results,time_zone,last_gtid