Cluster: configurable replicas migration barrier.

It is possible to configure the min number of additional working slaves
a master should be left with, for a slave to migrate to an orphaned
master.
This commit is contained in:
antirez 2014-01-31 11:12:34 +01:00
parent 3ff1bb4b2e
commit a7d30681c9
6 changed files with 39 additions and 2 deletions

View File

@ -551,6 +551,25 @@ lua-time-limit 5000
#
# cluster-node-timeout 15000
# Cluster slaves are able to migrate to orphaned masters, that are masters
# that are left without working slaves. This improves the cluster ability
# to resist to failures as otherwise an orphaned master can't be failed over
# in case of failure if it has no working slaves.
#
# Slaves migrate to orphaned masters only if there are still at least a
# given number of other working slaves for their old master. This number
# is the "migration barrer". A migration barrier of 1 means that a slave
# will migrate only if there is at least 1 other working slave for its master
# and so forth. It usually reflects the number of slaves you want for every
# master in your cluster.
#
# Default is 1 (slaves migrate only if their masters remain with at least
# one master). To disable migration just set it to a very large value.
# A value of 0 can be set but is useful only for debugging and dangerous
# in production.
#
# cluster-migration-barrier 1
# In order to setup your cluster make sure to read the documentation
# available at http://redis.io web site.

View File

@ -2141,12 +2141,13 @@ void clusterHandleSlaveMigration(int max_slaves) {
/* Step 1: Don't migrate if the cluster state is not ok. */
if (server.cluster->state != REDIS_CLUSTER_OK) return;
/* Step 2: Don't migrate if my master has just me as working slave. */
/* Step 2: Don't migrate if my master will not be left with at least
* 'migration-barrier' slaves after my migration. */
if (mymaster == NULL) return;
for (j = 0; j < mymaster->numslaves; j++)
if (!nodeFailed(mymaster->slaves[j]) &&
!nodeTimedOut(mymaster->slaves[j])) okslaves++;
if (okslaves == 1) return;
if (okslaves <= server.cluster_migration_barrier) return;
/* Step 3: Idenitfy a candidate for migration, and check if among the
* masters with the greatest number of ok slaves, I'm the one with the

View File

@ -21,6 +21,7 @@
#define REDIS_CLUSTER_SLAVE_VALIDITY_MULT 10 /* Slave data validity. */
#define REDIS_CLUSTER_FAILOVER_AUTH_RETRY_MULT 4 /* Auth request retry time. */
#define REDIS_CLUSTER_FAILOVER_DELAY 5 /* Seconds */
#define REDIS_CLUSTER_DEFAULT_MIGRATION_BARRIER 1
struct clusterNode;

View File

@ -422,6 +422,14 @@ void loadServerConfigFromString(char *config) {
if (server.cluster_node_timeout <= 0) {
err = "cluster node timeout must be 1 or greater"; goto loaderr;
}
} else if (!strcasecmp(argv[0],"cluster-migration-barrier")
&& argc == 2)
{
server.cluster_migration_barrier = atoi(argv[1]);
if (server.cluster_migration_barrier < 0) {
err = "cluster migration barrier must be positive";
goto loaderr;
}
} else if (!strcasecmp(argv[0],"lua-time-limit") && argc == 2) {
server.lua_time_limit = strtoll(argv[1],NULL,10);
} else if (!strcasecmp(argv[0],"slowlog-log-slower-than") &&
@ -875,6 +883,10 @@ void configSetCommand(redisClient *c) {
if (getLongLongFromObject(o,&ll) == REDIS_ERR ||
ll <= 0) goto badfmt;
server.cluster_node_timeout = ll;
} else if (!strcasecmp(c->argv[2]->ptr,"cluster-migration-barrier")) {
if (getLongLongFromObject(o,&ll) == REDIS_ERR ||
ll < 0) goto badfmt;
server.cluster_migration_barrier = ll;
} else {
addReplyErrorFormat(c,"Unsupported CONFIG parameter: %s",
(char*)c->argv[2]->ptr);
@ -975,6 +987,7 @@ void configGetCommand(redisClient *c) {
config_get_numerical_field("min-slaves-max-lag",server.repl_min_slaves_max_lag);
config_get_numerical_field("hz",server.hz);
config_get_numerical_field("cluster-node-timeout",server.cluster_node_timeout);
config_get_numerical_field("cluster-migration-barrier",server.cluster_migration_barrier);
/* Bool (yes/no) values */
config_get_bool_field("no-appendfsync-on-rewrite",
@ -1742,6 +1755,7 @@ int rewriteConfig(char *path) {
rewriteConfigYesNoOption(state,"cluster-enabled",server.cluster_enabled,0);
rewriteConfigStringOption(state,"cluster-config-file",server.cluster_configfile,REDIS_DEFAULT_CLUSTER_CONFIG_FILE);
rewriteConfigNumericalOption(state,"cluster-node-timeout",server.cluster_node_timeout,REDIS_CLUSTER_DEFAULT_NODE_TIMEOUT);
rewriteConfigNumericalOption(state,"cluster-migration-barrier",server.cluster_migration_barrier,REDIS_CLUSTER_DEFAULT_MIGRATION_BARRIER);
rewriteConfigNumericalOption(state,"slowlog-log-slower-than",server.slowlog_log_slower_than,REDIS_SLOWLOG_LOG_SLOWER_THAN);
rewriteConfigNumericalOption(state,"slowlog-max-len",server.slowlog_max_len,REDIS_SLOWLOG_MAX_LEN);
rewriteConfigNotifykeyspaceeventsOption(state);

View File

@ -1394,6 +1394,7 @@ void initServerConfig() {
server.repl_min_slaves_max_lag = REDIS_DEFAULT_MIN_SLAVES_MAX_LAG;
server.cluster_enabled = 0;
server.cluster_node_timeout = REDIS_CLUSTER_DEFAULT_NODE_TIMEOUT;
server.cluster_migration_barrier = REDIS_CLUSTER_DEFAULT_MIGRATION_BARRIER;
server.cluster_configfile = zstrdup(REDIS_DEFAULT_CLUSTER_CONFIG_FILE);
server.lua_caller = NULL;
server.lua_time_limit = REDIS_LUA_TIME_LIMIT;

View File

@ -791,6 +791,7 @@ struct redisServer {
mstime_t cluster_node_timeout; /* Cluster node timeout. */
char *cluster_configfile; /* Cluster auto-generated config file name. */
struct clusterState *cluster; /* State of the cluster */
int cluster_migration_barrier; /* Cluster replicas migration barrier. */
/* Scripting */
lua_State *lua; /* The Lua interpreter. We use just one for all clients */
redisClient *lua_client; /* The "fake client" to query Redis from Lua */