hcq/MaxScale
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Added 'disable_master_failback' monitor option

Browse Source 
Added 'disable_master_failback' option in Galera monitor
This commit is contained in:
MassimilianoPinto
2014-10-30 19:03:07 +01:00
parent 9967a45379
commit 6f22975e6c
7 changed files with 64 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

44
server/modules/monitor/galera_mon.c
View File

@ -30,7 +30,7 @@
* Interval is printed in diagnostics.
* 03/06/14 Mark Riddoch Add support for maintenance mode
* 24/06/14 Massimiliano Pinto Added depth level 0 for each node
* 30/10/14 Massimiliano Pinto Added stickininess for master selection
* 30/10/14 Massimiliano Pinto Added disableMasterFailback feature
*
* @endverbatim
*/
@ -53,7 +53,7 @@ extern int lm_enabled_logfiles_bitmask;
static void monitorMain(void *);
static char *version_str = "V1.2.0";
static char *version_str = "V1.3.0";
MODULE_INFO info = {
MODULE_API_MONITOR,
@ -71,6 +71,7 @@ static void diagnostics(DCB *, void *);
static void setInterval(void *, size_t);
static MONITOR_SERVERS *get_candidate_master(MONITOR_SERVERS *);
static MONITOR_SERVERS *set_cluster_master(MONITOR_SERVERS *, MONITOR_SERVERS *, int);
static void disableMasterFailback(void *, int);
static MONITOR_OBJECT MyObject = {
startMonitor,
@ -83,6 +84,7 @@ static MONITOR_OBJECT MyObject = {
NULL,
NULL,
NULL,
disableMasterFailback
};
/**
@ -150,6 +152,7 @@ MYSQL_MONITOR *handle;
handle->defaultPasswd = NULL;
handle->id = MONITOR_DEFAULT_ID;
handle->interval = MONITOR_INTERVAL;
handle->disableMasterFailback = 0;
handle->master = NULL;
spinlock_init(&handle->lock);
}
@ -435,11 +438,11 @@ char *server_string;
static void
monitorMain(void *arg)
{
MYSQL_MONITOR *handle = (MYSQL_MONITOR *)arg;
MONITOR_SERVERS *ptr;
size_t nrounds = 0;
MONITOR_SERVERS *candidate_master=NULL;
int master_stickiness=1;
MYSQL_MONITOR *handle = (MYSQL_MONITOR *)arg;
MONITOR_SERVERS *ptr;
size_t nrounds = 0;
MONITOR_SERVERS *candidate_master = NULL;
int master_stickiness = handle->disableMasterFailback;
if (mysql_thread_init())
{
@ -486,7 +489,6 @@ int master_stickiness=1;
/* clear bits for non member nodes */
if ( ! SERVER_IN_MAINT(ptr->server) && (ptr->server->node_id < 0 || ! SERVER_IS_JOINED(ptr->server))) {
ptr->server->depth = -1;
/* clear M/S status */
@ -525,7 +527,6 @@ int master_stickiness=1;
/* Select the master, based on master_stickiness */
handle->master = set_cluster_master(handle->master, candidate_master, master_stickiness);
ptr = handle->databases;
while (ptr && handle->master) {
@ -573,16 +574,14 @@ MYSQL_MONITOR *handle = (MYSQL_MONITOR *)arg;
memcpy(&handle->interval, &interval, sizeof(unsigned long));
}
/**
* get candidate master from all nodes
*
* current available rule: get the server with MIN(node_id)
* The current available rule: get the server with MIN(node_id)
* node_id comes from 'wsrep_local_index' variable
*
* @param servers The monitored servers list
* @return The candidate master on success, NULL on failure
*
*/
static MONITOR_SERVERS *get_candidate_master(MONITOR_SERVERS *servers) {
MONITOR_SERVERS *ptr = servers;
@ -618,7 +617,7 @@ static MONITOR_SERVERS *get_candidate_master(MONITOR_SERVERS *servers) {
* The selection is based on the configuration option mapped to master_stickiness
* The candidate master may change over time due to
* 'wsrep_local_index' value change in the Galera Cluster
* Enablig master_stickiness will avoid master change unless a failure is spotted
* Enabling master_stickiness will avoid master change unless a failure is spotted
*
* @param current_master Previous master server
* @param candidate_master The candidate master server accordingly to the selection rule
@ -642,3 +641,22 @@ static MONITOR_SERVERS *set_cluster_master(MONITOR_SERVERS *current_master, MONI
return candidate_master;
}
}
/**
* Disable/Enable the Master failback in a Galera Cluster.
*
* A restarted / rejoined node may get back the previous 'wsrep_local_index'
* from Cluster: if the value is the lowest in the cluster it will be selected as Master
* This will cause a Master change even if there is no failure.
* The option if set to 1 will avoid this situation, keeping the current Master (if running) available
*
* @param arg The handle allocated by startMonitor
* @param disable To disable it use 1, 0 keeps failback
*/
static void
disableMasterFailback(void *arg, int disable)
{
MYSQL_MONITOR *handle = (MYSQL_MONITOR *)arg;
memcpy(&handle->disableMasterFailback, &disable, sizeof(int));
}