Merge branch 'develop' into MAX-11

Conflicts:
	server/include/server.h
	server/modules/monitor/mysql_mon.c
	server/modules/protocol/mysql_client.c
	server/modules/routing/readwritesplit/readwritesplit.c
	utils/skygw_debug.h
This commit is contained in:
VilhoRaatikka
2014-06-07 00:50:08 +03:00
47 changed files with 2759 additions and 128 deletions

View File

@ -30,6 +30,9 @@
* diagnostic interface
* 20/05/14 Massimiliano Pinto Addition of support for MariadDB multimaster replication setup.
* New server field version_string is updated.
* 28/05/14 Massimiliano Pinto Added set Id and configuration options (setInverval)
* Parameters are now printed in diagnostics
* 03/06/14 Mark Ridoch Add support for maintenance mode
*
* @endverbatim
*/
@ -46,12 +49,20 @@
#include <log_manager.h>
#include <secrets.h>
#include <dcb.h>
#include <modinfo.h>
extern int lm_enabled_logfiles_bitmask;
static void monitorMain(void *);
static char *version_str = "V1.1.0";
static char *version_str = "V1.2.0";
MODULE_INFO info = {
MODULE_API_MONITOR,
MODULE_ALPHA_RELEASE,
MONITOR_VERSION,
"A MySQL Master/Slave replication monitor"
};
static void *startMonitor(void *);
static void stopMonitor(void *);
@ -59,8 +70,11 @@ static void registerServer(void *, SERVER *);
static void unregisterServer(void *, SERVER *);
static void defaultUser(void *, char *, char *);
static void diagnostics(DCB *, void *);
static void setInterval(void *, unsigned long);
static void defaultId(void *, unsigned long);
static void replicationHeartbeat(void *, int);
static MONITOR_OBJECT MyObject = { startMonitor, stopMonitor, registerServer, unregisterServer, defaultUser, diagnostics };
static MONITOR_OBJECT MyObject = { startMonitor, stopMonitor, registerServer, unregisterServer, defaultUser, diagnostics, setInterval, defaultId, replicationHeartbeat };
/**
* Implementation of the mandatory version entry point
@ -126,6 +140,8 @@ MYSQL_MONITOR *handle;
handle->shutdown = 0;
handle->defaultUser = NULL;
handle->defaultPasswd = NULL;
handle->id = MONITOR_DEFAULT_ID;
handle->interval = MONITOR_INTERVAL;
spinlock_init(&handle->lock);
}
handle->tid = (THREAD)thread_start(monitorMain, handle);
@ -261,7 +277,12 @@ char *sep;
dcb_printf(dcb, "\tMonitor stopped\n");
break;
}
dcb_printf(dcb,"\tSampling interval:\t%lu milliseconds\n", handle->interval);
dcb_printf(dcb,"\tMaxScale MonitorId:\t%lu\n", handle->id);
dcb_printf(dcb,"\tReplication lag:\t%s\n", (handle->replicationHeartbeat == 1) ? "enabled" : "disabled");
dcb_printf(dcb, "\tMonitored servers: ");
db = handle->databases;
sep = "";
while (db)
@ -280,13 +301,13 @@ char *sep;
/**
* Monitor an individual server
*
* @param handle The MySQL Monitor object
* @param database The database to probe
* @param defaultUser Default username for the monitor
* @param defaultPasswd Default password for the monitor
*/
static void
monitorDatabase(MONITOR_SERVERS *database, char *defaultUser, char *defaultPasswd)
monitorDatabase(MYSQL_MONITOR *handle, MONITOR_SERVERS *database)
{
<<<<<<< HEAD
MYSQL_ROW row;
MYSQL_RES *result;
int num_fields;
@ -298,6 +319,21 @@ static int conn_err_count;
static int modval = 10;
if (database->server->monuser != NULL)
=======
MYSQL_ROW row;
MYSQL_RES *result;
int num_fields;
int ismaster = 0, isslave = 0;
char *uname = handle->defaultUser, *passwd = handle->defaultPasswd;
unsigned long int server_version = 0;
char *server_string;
unsigned long id = handle->id;
int replication_heartbeat = handle->replicationHeartbeat;
static int conn_err_count;
static int modval = 10;
if (database->server->monuser != NULL)
>>>>>>> develop
{
uname = database->server->monuser;
passwd = database->server->monpw;
@ -306,13 +342,24 @@ static int modval = 10;
if (uname == NULL)
return;
<<<<<<< HEAD
=======
/* Don't probe servers in maintenance mode */
if (SERVER_IN_MAINT(database->server))
return;
>>>>>>> develop
if (database->con == NULL || mysql_ping(database->con) != 0)
{
char *dpwd = decryptPassword(passwd);
int rc;
int read_timeout = 1;
<<<<<<< HEAD
database->con = mysql_init(NULL);
=======
database->con = mysql_init(NULL);
>>>>>>> develop
rc = mysql_options(database->con, MYSQL_OPT_READ_TIMEOUT, (void *)&read_timeout);
if (mysql_real_connect(database->con,
@ -324,6 +371,7 @@ static int modval = 10;
NULL,
0) == NULL)
{
<<<<<<< HEAD
if (conn_err_count%modval == 0)
{
LOGIF(LE, (skygw_log_write_flush(
@ -340,6 +388,16 @@ static int modval = 10;
{
conn_err_count += 1;
}
=======
LOGIF(LE, (skygw_log_write_flush(
LOGFILE_ERROR,
"Error : Monitor was unable to connect to "
"server %s:%d : \"%s\"",
database->server->name,
database->server->port,
mysql_error(database->con))));
>>>>>>> develop
free(dpwd);
server_clear_status(database->server, SERVER_RUNNING);
@ -386,15 +444,112 @@ static int modval = 10;
{
/* Log lack of permission */
}
}
else if ((result = mysql_store_result(database->con)) != NULL)
{
database->server->rlag = -1;
} else if ((result = mysql_store_result(database->con)) != NULL) {
num_fields = mysql_num_fields(result);
while ((row = mysql_fetch_row(result)))
{
ismaster = 1;
}
mysql_free_result(result);
if (ismaster && replication_heartbeat == 1) {
time_t heartbeat;
time_t purge_time;
char heartbeat_insert_query[128]="";
char heartbeat_purge_query[128]="";
handle->master_id = database->server->node_id;
/* create the maxscale_schema database */
if (mysql_query(database->con, "CREATE DATABASE IF NOT EXISTS maxscale_schema")) {
LOGIF(LE, (skygw_log_write_flush(
LOGFILE_ERROR,
"[mysql_mon]: Error creating maxscale_schema database in Master server"
": %s", mysql_error(database->con))));
database->server->rlag = -1;
}
/* create repl_heartbeat table in maxscale_schema database */
if (mysql_query(database->con, "CREATE TABLE IF NOT EXISTS "
"maxscale_schema.replication_heartbeat "
"(maxscale_id INT NOT NULL, "
"master_server_id INT NOT NULL, "
"master_timestamp INT UNSIGNED NOT NULL, "
"PRIMARY KEY ( master_server_id, maxscale_id ) ) "
"ENGINE=MYISAM DEFAULT CHARSET=latin1")) {
LOGIF(LE, (skygw_log_write_flush(
LOGFILE_ERROR,
"[mysql_mon]: Error creating maxscale_schema.replication_heartbeat table in Master server"
": %s", mysql_error(database->con))));
database->server->rlag = -1;
}
/* auto purge old values after 48 hours*/
purge_time = time(0) - (3600 * 48);
sprintf(heartbeat_purge_query, "DELETE FROM maxscale_schema.replication_heartbeat WHERE master_timestamp < %lu", purge_time);
if (mysql_query(database->con, heartbeat_purge_query)) {
LOGIF(LE, (skygw_log_write_flush(
LOGFILE_ERROR,
"[mysql_mon]: Error deleting from maxscale_schema.replication_heartbeat table: [%s], %s",
heartbeat_purge_query,
mysql_error(database->con))));
}
heartbeat = time(0);
/* set node_ts for master as time(0) */
database->server->node_ts = heartbeat;
sprintf(heartbeat_insert_query, "UPDATE maxscale_schema.replication_heartbeat SET master_timestamp = %lu WHERE master_server_id = %i AND maxscale_id = %lu", heartbeat, handle->master_id, id);
/* Try to insert MaxScale timestamp into master */
if (mysql_query(database->con, heartbeat_insert_query)) {
database->server->rlag = -1;
LOGIF(LE, (skygw_log_write_flush(
LOGFILE_ERROR,
"[mysql_mon]: Error updating maxscale_schema.replication_heartbeat table: [%s], %s",
heartbeat_insert_query,
mysql_error(database->con))));
} else {
if (mysql_affected_rows(database->con) == 0) {
heartbeat = time(0);
sprintf(heartbeat_insert_query, "REPLACE INTO maxscale_schema.replication_heartbeat (master_server_id, maxscale_id, master_timestamp ) VALUES ( %i, %lu, %lu)", handle->master_id, id, heartbeat);
if (mysql_query(database->con, heartbeat_insert_query)) {
database->server->rlag = -1;
LOGIF(LE, (skygw_log_write_flush(
LOGFILE_ERROR,
"[mysql_mon]: Error inserting into maxscale_schema.replication_heartbeat table: [%s], %s",
heartbeat_insert_query,
mysql_error(database->con))));
} else {
/* Set replication lag to 0 for the master */
database->server->rlag = 0;
LOGIF(LD, (skygw_log_write_flush(
LOGFILE_DEBUG,
"[mysql_mon]: heartbeat table inserted data for %s:%i", database->server->name, database->server->port)));
}
} else {
/* Set replication lag as 0 for the master */
database->server->rlag = 0;
LOGIF(LD, (skygw_log_write_flush(
LOGFILE_DEBUG,
"[mysql_mon]: heartbeat table updated for %s:%i", database->server->name, database->server->port)));
}
}
}
}
/* Check if the Slave_SQL_Running and Slave_IO_Running status is
@ -439,6 +594,80 @@ static int modval = 10;
}
}
/* Get the master_timestamp value from maxscale_schema.replication_heartbeat table */
if (isslave && replication_heartbeat == 1) {
time_t heartbeat;
char select_heartbeat_query[256] = "";
sprintf(select_heartbeat_query, "SELECT master_timestamp "
"FROM maxscale_schema.replication_heartbeat "
"WHERE maxscale_id = %lu AND master_server_id = %i",
id, handle->master_id);
/* if there is a master then send the query to the slave with master_id*/
if (handle->master_id >= 0 && (mysql_query(database->con, select_heartbeat_query) == 0
&& (result = mysql_store_result(database->con)) != NULL)) {
num_fields = mysql_num_fields(result);
while ((row = mysql_fetch_row(result))) {
int rlag = -1;
time_t slave_read;
heartbeat = time(0);
slave_read = strtoul(row[0], NULL, 10);
if ((errno == ERANGE && (slave_read == LONG_MAX || slave_read == LONG_MIN)) || (errno != 0 && slave_read == 0)) {
slave_read = 0;
}
if (slave_read) {
/* set the replication lag */
rlag = heartbeat - slave_read;
}
/* set this node_ts as master_timestamp read from replication_heartbeat table */
database->server->node_ts = slave_read;
if (rlag >= 0) {
/* store rlag only if greater than monitor sampling interval */
database->server->rlag = (rlag > (handle->interval / 1000)) ? rlag : 0;
} else {
database->server->rlag = -1;
}
LOGIF(LD, (skygw_log_write_flush(
LOGFILE_DEBUG,
"[mysql_mon]: replication heartbeat: "
"server %s:%i is %i seconds behind master",
database->server->name,
database->server->port,
database->server->rlag)));
}
mysql_free_result(result);
} else {
database->server->rlag = -1;
database->server->node_ts = 0;
if (handle->master_id < 0) {
LOGIF(LE, (skygw_log_write_flush(
LOGFILE_ERROR,
"[mysql_mon]: error: replication heartbeat: "
"master_server_id NOT available for %s:%i",
database->server->name,
database->server->port)));
} else {
LOGIF(LE, (skygw_log_write_flush(
LOGFILE_ERROR,
"[mysql_mon]: error: replication heartbeat: "
"failed selecting from hearthbeat table of %s:%i : [%s], %s",
database->server->name,
database->server->port,
select_heartbeat_query,
mysql_error(database->con))));
}
}
}
if (ismaster)
{
server_set_status(database->server, SERVER_MASTER);
@ -492,11 +721,18 @@ static int modval = 10;
{
unsigned int prev_status = ptr->server->status;
<<<<<<< HEAD
monitorDatabase(ptr, handle->defaultUser, handle->defaultPasswd);
if (ptr->server->status != prev_status ||
(SERVER_IS_DOWN(ptr->server) &&
err_count%modval == 0))
=======
monitorDatabase(handle, ptr);
if (ptr->server->status != prev_status ||
SERVER_IS_DOWN(ptr->server))
>>>>>>> develop
{
LOGIF(LM, (skygw_log_write_flush(
LOGFILE_MESSAGE,
@ -504,6 +740,7 @@ static int modval = 10;
ptr->server->name,
ptr->server->port,
STRSRVSTATUS(ptr->server))));
<<<<<<< HEAD
err_count = 0;
modval += 1;
}
@ -511,8 +748,51 @@ static int modval = 10;
{
err_count += 1;
}
=======
}
>>>>>>> develop
ptr = ptr->next;
}
thread_millisleep(10000);
thread_millisleep(handle->interval);
}
}
/**
* Set the default id to use in the monitor.
*
* @param arg The handle allocated by startMonitor
* @param id The id to set in monitor struct
*/
static void
defaultId(void *arg, unsigned long id)
{
MYSQL_MONITOR *handle = (MYSQL_MONITOR *)arg;
memcpy(&handle->id, &id, sizeof(unsigned long));
}
/**
* Set the monitor sampling interval.
*
* @param arg The handle allocated by startMonitor
* @param interval The interval to set in monitor struct, in milliseconds
*/
static void
setInterval(void *arg, unsigned long interval)
{
MYSQL_MONITOR *handle = (MYSQL_MONITOR *)arg;
memcpy(&handle->interval, &interval, sizeof(unsigned long));
}
/**
* Enable/Disable the MySQL Replication hearbeat, detecting slave lag behind master.
*
* @param arg The handle allocated by startMonitor
* @param replicationHeartbeat To enable it 1, disable it with 0
*/
static void
replicationHeartbeat(void *arg, int replicationHeartbeat)
{
MYSQL_MONITOR *handle = (MYSQL_MONITOR *)arg;
memcpy(&handle->replicationHeartbeat, &replicationHeartbeat, sizeof(int));
}