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

Pick node with index of 0 as the Galera master

Browse Source 
When a Galera cluster loses a member, it will recalculate the
wsrep_local_index values. As the index is zero-based, we can be certain
that in a valid cluster there will always be a node with an index of 0.

If the galeramon can't find a node with an index of 0, it means that
either the cluster hasn't stabilized and there's a pending recalculation
of the index or that there's no connectivity between MaxScale and the node
with the index value 0.

With this change and default settings, active-active MaxScale setups with
Galera clusters should always choose the same node as the master.
This commit is contained in:
Markus Mäkelä 2016-12-16 19:33:05 +02:00
parent 067d48ddc6
commit 2b5ec8f162
4 changed files with 55 additions and 16 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
Documentation/Monitors/Galera-Monitor.md
View File

@ -62,6 +62,20 @@ Enable interaction with server priorities. This will allow the monitor to determ
use_priority=true
```
### `root_node_as_master`
This option controls whether the write master Galera node requires a
_wsrep_local_index_ value of 0. This option is enabled by default and was
introduced in MaxScale 2.1.0.
A Galera cluster will always have a node which has a _wsrep_local_index_ value
of 0. Based on this information, multiple MaxScale instances can always pick the
same node for writes.
If the `root_node_as_master` option is disabled for galeramon, the node with the
lowest index will always be chosen as the master. If it is enabled, only the
node with a a _wsrep_local_index_ value of 0 can be chosed as the master.
## Interaction with Server Priorities
If the `use_priority` option is set and a server is configured with the

9
Documentation/Release-Notes/MaxScale-2.1.0-Release-Notes.md
View File

@ -94,6 +94,15 @@ In 2.1.0, each listener has its own sub-directory in the service cache
directory. The old caches in `/var/cache/maxscale` will need to be manually
removed if they are no longer used by older versions of MaxScale.
### Galeramon Monitoring Algorithm
The galeramon monitor will only choose nodes with a _wsrep_local_index_
value of 0 as the master. This allows multiple MaxScales to always choose
the same node as the write master node for the cluster. The old behavior
can be taken into use by disabling the new `root_node_as_master` option.
For more details, read the [Galeramon documentation](../Monitors/Galera-Monitor.md).
## New Features
### Dynamic server configuration

46
server/modules/monitor/galeramon/galeramon.c
View File

@ -145,6 +145,7 @@ startMonitor(MONITOR *mon, const CONFIG_PARAMETER *params)
handle->disableMasterRoleSetting = 0;
handle->master = NULL;
handle->script = NULL;
handle->root_node_as_master = true;
handle->use_priority = false;
memset(handle->events, false, sizeof(handle->events));
spinlock_init(&handle->lock);
@ -165,6 +166,10 @@ startMonitor(MONITOR *mon, const CONFIG_PARAMETER *params)
{
handle->disableMasterRoleSetting = config_truth_value(params->value);
}
else if (!strcmp(params->name, "root_node_as_master"))
{
handle->root_node_as_master = config_truth_value(params->value);
}
else if (!strcmp(params->name, "use_priority"))
{
handle->use_priority = config_truth_value(params->value);
@ -548,29 +553,27 @@ monitorMain(void *arg)
const int repl_bits = (SERVER_SLAVE | SERVER_MASTER | SERVER_MASTER_STICKINESS);
if (SERVER_IS_JOINED(ptr->server))
{
if (handle->master)
if (ptr != handle->master)
{
if (ptr != handle->master)
/* set the Slave role and clear master stickiness */
server_clear_set_status(ptr->server, repl_bits, SERVER_SLAVE);
}
else
{
if (candidate_master &&
handle->master->server->node_id != candidate_master->server->node_id)
{
/* set the Slave role and clear master stickiness */
server_clear_set_status(ptr->server, repl_bits, SERVER_SLAVE);
/* set master role and master stickiness */
server_clear_set_status(ptr->server, repl_bits,
(SERVER_MASTER | SERVER_MASTER_STICKINESS));
}
else
{
if (candidate_master &&
handle->master->server->node_id != candidate_master->server->node_id)
{
/* set master role and master stickiness */
server_clear_set_status(ptr->server, repl_bits,
(SERVER_MASTER | SERVER_MASTER_STICKINESS));
}
else
{
/* set master role and clear master stickiness */
server_clear_set_status(ptr->server, repl_bits, SERVER_MASTER);
}
/* set master role and clear master stickiness */
server_clear_set_status(ptr->server, repl_bits, SERVER_MASTER);
}
}
is_cluster++;
}
else
@ -674,6 +677,17 @@ static MONITOR_SERVERS *get_candidate_master(MONITOR* mon)
moitor_servers = moitor_servers->next;
}
if (!handle->use_priority && !handle->disableMasterFailback &&
handle->root_node_as_master && min_id > 0)
{
/** The monitor couldn't find the node with wsrep_local_index of 0.
* This means that we can't connect to the root node of the cluster.
*
* If the node is down, the cluster would recalculate the index values
* and we would find it. In this case, we just can't connect to it. */
candidate_master = NULL;
}
return candidate_master;
}

2
server/modules/monitor/galeramon/galeramon.h
View File

@ -58,6 +58,8 @@ typedef struct
int disableMasterRoleSetting; /**< Monitor flag to disable setting master role */
MONITOR_SERVERS *master; /**< Master server for MySQL Master/Slave replication */
char* script;
bool root_node_as_master; /**< Whether we require that the Master should
* have a wsrep_local_index of 0 */
bool use_priority; /*< Use server priorities */
bool events[MAX_MONITOR_EVENT]; /*< enabled events */
} GALERA_MONITOR;