pgpool2.git / summary / log / commit / refs
commit 07c767dc0dd71d959efd97772ccc07b4f40a81e0
Author: Muhammad Usama <m.usama@gmail.com>
Date: Mon Dec 07 15:22:06 2015 +0000
code cleanup for watchdog.
doc/pgpool-en.html | 3 +-
src/include/watchdog/watchdog.h | 25 +-
src/include/watchdog/wd_ipc_commands.h | 2 +-
src/include/watchdog/wd_ipc_defines.h | 18 +-
src/main/pgpool_main.c | 2 +-
src/tools/pcp/pcp_frontend_client.c | 4 +-
src/watchdog/watchdog.c | 228 +------------------
src/watchdog/wd_commands.c | 132 +++++------
src/watchdog/wd_escalation.c | 44 ++--
src/watchdog/wd_interlock.c | 150 ------------
src/watchdog/wd_lifecheck.c | 15 +-
src/watchdog/wd_list.c | 403 ---------------------------------
src/watchdog/wd_ping.c | 1 -
13 files changed, 113 insertions(+), 914 deletions(-)
diff --git a/doc/pgpool-en.html b/doc/pgpool-en.html
index b3d9a16c9..7ffa1d086 100644
--- a/doc/pgpool-en.html
+++ b/doc/pgpool-en.html
@@ -4534,12 +4534,11 @@ string and the socket file is placed in the pgpool-II <a href="#WD_IPC_SOCKET_DI
<h3 id="Watchdog_IPC_command_packet_format">Watchdog IPC command packet format</h3>
<dl>
<p>
-The watchdog IPC command packet consists of four fields. Below table details the message fields and description
+The watchdog IPC command packet consists of three fields. Below table details the message fields and description
</p>
<table border>
<tr class="header"><th>Field</th> <th>Type</th> <th>Description</th></tr>
<tr><th>TYPE</th> <td>BYTE1</td> <td>Command Type</td></tr>
- <tr><th>ACTION</th> <td>INT32 Network byte order</td> <td>Command Action. Currently only DEFAULT(0) is supported</td></tr>
<tr><th>LENGTH</th> <td>INT32 Network byte order</td> <td>The length of data to follow</td></tr>
<tr><th>DATA</th> <td>DATA in JSON format</td> <td>Command data in JSON format</td></tr>
</table>
diff --git a/src/include/watchdog/watchdog.h b/src/include/watchdog/watchdog.h
index bf0169f77..0132d31ac 100644
--- a/src/include/watchdog/watchdog.h
+++ b/src/include/watchdog/watchdog.h
@@ -87,32 +87,27 @@ typedef enum {
typedef struct SocketConnection
{
- int sock;
- struct timeval tv;
- char addr[48];
- WD_SOCK_STATE sock_state;
+ int sock; /* socket descriptor */
+ struct timeval tv; /* connect time of socket */
+ char addr[48]; /* ip address of socket connection*/
+ WD_SOCK_STATE sock_state; /* current state of socket */
}SocketConnection;
typedef struct WatchdogNode
{
WD_STATES state;
struct timeval tv; /* startup time value */
- char nodeName[WD_MAX_HOST_NAMELEN];
+ char nodeName[WD_MAX_HOST_NAMELEN]; /* name of this node */
char hostname[WD_MAX_HOST_NAMELEN]; /* host name */
int wd_port; /* watchdog port */
int pgpool_port; /* pgpool port */
int wd_priority; /* watchdog priority */
char delegate_ip[WD_MAX_HOST_NAMELEN]; /* delegate IP */
- char** resolved_ips;
- int delegate_ip_flag; /* delegate IP flag */
- unsigned int lastCommandID;
- struct timeval hb_last_recv_time; /* recv time */
- int private_id;
- SocketConnection server_socket;
- SocketConnection client_socket;
- bool is_connectable; /* true if any of the socket is connected */
- bool is_lock_holder; /* lock holder flag */
- bool in_interlocking; /* interlocking is in progress */
+ int private_id; /* ID assigned to this node
+ * This id is consumed locally
+ */
+ SocketConnection server_socket; /* socket connections for this node initiated by remote */
+ SocketConnection client_socket; /* socket connections for this node initiated by local*/
}WatchdogNode;
extern pid_t initialize_watchdog(void);
diff --git a/src/include/watchdog/wd_ipc_commands.h b/src/include/watchdog/wd_ipc_commands.h
index f71bc1980..2e969c30e 100644
--- a/src/include/watchdog/wd_ipc_commands.h
+++ b/src/include/watchdog/wd_ipc_commands.h
@@ -64,7 +64,7 @@ extern WDFailoverCMDResults wd_send_failover_sync_command(WDFailoverCMDTypes cmd
extern char* wd_get_watchdog_nodes(int nodeID);
-extern WDIPCCmdResult* issue_command_to_watchdog(char type, WD_COMMAND_ACTIONS command_action,int timeout_sec, char* data, int data_len, bool blocking);
+extern WDIPCCmdResult* issue_command_to_watchdog(char type, int timeout_sec, char* data, int data_len, bool blocking);
/* wd_interlock.c */
diff --git a/src/include/watchdog/wd_ipc_defines.h b/src/include/watchdog/wd_ipc_defines.h
index 4b7b12d42..2c3c6b661 100644
--- a/src/include/watchdog/wd_ipc_defines.h
+++ b/src/include/watchdog/wd_ipc_defines.h
@@ -6,7 +6,7 @@
* pgpool: a language independent connection pool server for PostgreSQL
* written by Tatsuo Ishii
*
- * Copyright (c) 2003-2012 PgPool Global Development Group
+ * Copyright (c) 2003-2015 PgPool Global Development Group
*
* Permission to use, copy, modify, and distribute this software and
* its documentation for any purpose and without fee is hereby
@@ -42,13 +42,6 @@ typedef enum WDFailoverCMDResults
FAILOVER_RES_BLOCKED
}WDFailoverCMDResults;
-typedef enum {
- WD_COMMAND_ACTION_DEFAULT = 0,
- WD_COMMAND_ACTION_SEND_ALL,
- WD_COMMAND_ACTION_SEND_MASTER,
- WD_COMMAND_ACTION_LOCAL
-} WD_COMMAND_ACTIONS;
-
/* IPC MESSAGES TYPES */
#define WD_REGISTER_FOR_NOTIFICATION '0'
@@ -56,12 +49,9 @@ typedef enum {
#define WD_GET_NODES_LIST_COMMAND '3'
#define WD_NODES_LIST_DATA '4'
-#define WD_TRY_COMMAND_LOCK '5'
-#define WD_COMMAND_UNLOCK '6'
-
-#define WD_IPC_CMD_CLUSTER_IN_TRAN '7'
-#define WD_IPC_CMD_RESULT_BAD '8'
-#define WD_IPC_CMD_RESULT_OK '9'
+#define WD_IPC_CMD_CLUSTER_IN_TRAN '5'
+#define WD_IPC_CMD_RESULT_BAD '6'
+#define WD_IPC_CMD_RESULT_OK '7'
#define WD_FUNCTION_COMMAND 'f'
#define WD_FAILOVER_CMD_SYNC_REQUEST 's'
diff --git a/src/main/pgpool_main.c b/src/main/pgpool_main.c
index ea766034f..6dbe16119 100644
--- a/src/main/pgpool_main.c
+++ b/src/main/pgpool_main.c
@@ -395,7 +395,7 @@ int PgpoolMain(bool discard_status, bool clear_memcache_oidmaps)
if (ret > 0) /* Retries are exhausted, reset the counter */
{
retrycnt = 0;
- if (ret == 2)
+ if (ret == 2) /* 2 = failover done on node */
{
health_check_node_id = 0;
use_template_db = false;
diff --git a/src/tools/pcp/pcp_frontend_client.c b/src/tools/pcp/pcp_frontend_client.c
index a4ee713bf..018c80bfc 100644
--- a/src/tools/pcp/pcp_frontend_client.c
+++ b/src/tools/pcp/pcp_frontend_client.c
@@ -218,7 +218,7 @@ main(int argc, char **argv)
{
if (nodeID < 0)
{
- fprintf(stderr, "%s: Invalid watchdog-id \"%s\", must be a positive number\n",progname,optarg);
+ fprintf(stderr, "%s: Invalid watchdog-id \"%s\", must be a positive number or zero for a local watchdog node\n",progname,optarg);
exit(0);
}
}
@@ -287,7 +287,7 @@ main(int argc, char **argv)
nodeID = atoi(argv[optind]);
if (nodeID < 0 )
{
- fprintf(stderr, "%s: Invalid watchdog-id \"%s\", must be a positive number\n",progname,optarg);
+ fprintf(stderr, "%s: Invalid watchdog-id \"%s\", must be a positive number or zero for local watchdog node\n",progname,optarg);
exit(0);
}
}
diff --git a/src/watchdog/watchdog.c b/src/watchdog/watchdog.c
index 6fe57b00d..9492dfc12 100644
--- a/src/watchdog/watchdog.c
+++ b/src/watchdog/watchdog.c
@@ -131,7 +131,7 @@ char *wd_event_name[] =
"NODE CONNECTION FOUND"
};
-char *debug_states[] = {
+char *wd_state_names[] = {
"DEAD",
"LOADING",
"JOINING",
@@ -175,7 +175,6 @@ typedef struct WDCommandNodeResult
typedef struct WDIPCCommandData
{
- WD_COMMAND_ACTIONS command_action;
MemoryContext memoryContext;
int issueing_sock;
char type;
@@ -267,9 +266,6 @@ typedef struct wd_cluster
List *wd_timer_commands;
}wd_cluster;
-int *escalation_status = NULL; /* Lives in shared memory */
-
-
volatile sig_atomic_t reload_config_signal = 0;
static void check_config_reload(void);
@@ -365,9 +361,7 @@ static WDIPCCommandData* get_wd_IPC_command_from_socket(int sock);
static IPC_CMD_PREOCESS_RES process_IPC_command(WDIPCCommandData* IPCCommand);
static IPC_CMD_PREOCESS_RES process_IPC_nodeStatusChange_command(WDIPCCommandData* IPCCommand);
-static IPC_CMD_PREOCESS_RES process_IPC_lock_request(WDIPCCommandData *IPCCommand);
static IPC_CMD_PREOCESS_RES process_IPC_nodeList_command(WDIPCCommandData* IPCCommand);
-static IPC_CMD_PREOCESS_RES process_IPC_unlock_request(WDIPCCommandData *IPCCommand);
static IPC_CMD_PREOCESS_RES process_IPC_replicate_variable(WDIPCCommandData* IPCCommand);
static IPC_CMD_PREOCESS_RES process_IPC_failover_cmd_synchronise(WDIPCCommandData *IPCCommand);
static IPC_CMD_PREOCESS_RES execute_replicate_command(WDIPCCommandData* ipcCommand);
@@ -491,9 +485,6 @@ static void wd_cluster_initialize(void)
(errmsg("watchdog remote node:%d on %s:%d",i,g_cluster.remoteNodes[i].hostname, g_cluster.remoteNodes[i].wd_port)));
}
- escalation_status = pool_shared_memory_create(sizeof(int));
- *escalation_status = 0;
-
g_cluster.masterNode = NULL;
g_cluster.aliveNodeCount = 0;
g_cluster.quorum_status = -1;
@@ -1338,7 +1329,6 @@ static bool write_ipc_command_with_result_data(WDIPCCommandData* IPCCommand, cha
static bool read_ipc_command_and_process(int sock, bool *remove_socket)
{
char type;
- WD_COMMAND_ACTIONS command_action;
IPC_CMD_PREOCESS_RES res;
int data_len,ret;
WDIPCCommandData* IPCCommand = NULL;
@@ -1357,16 +1347,7 @@ static bool read_ipc_command_and_process(int sock, bool *remove_socket)
errdetail("read from socket failed with error \"%s\"",strerror(errno))));
return false;
}
- /* Next is is command action */
- ret = socket_read(sock, &command_action, sizeof(int),0);
- if (ret != sizeof(int))
- {
- ereport(WARNING,
- (errmsg("error reading from IPC socket"),
- errdetail("read from socket failed with error \"%s\"",strerror(errno))));
- return false;
- }
- command_action = htonl(command_action);
+
/* We should have data length */
ret = socket_read(sock, &data_len, sizeof(int),0);
if (ret != sizeof(int))
@@ -1390,7 +1371,6 @@ static bool read_ipc_command_and_process(int sock, bool *remove_socket)
IPCCommand = palloc0(sizeof(WDIPCCommandData));
IPCCommand->issueing_sock = sock;
- IPCCommand->command_action = command_action;
IPCCommand->type = type;
gettimeofday(&IPCCommand->issue_time, NULL);
@@ -1446,14 +1426,6 @@ static IPC_CMD_PREOCESS_RES process_IPC_command(WDIPCCommandData* IPCCommand)
return process_IPC_nodeStatusChange_command(IPCCommand);
break;
- case WD_TRY_COMMAND_LOCK:
- return process_IPC_lock_request(IPCCommand);
- break;
-
- case WD_COMMAND_UNLOCK:
- process_IPC_unlock_request(IPCCommand);
- break;
-
case WD_REGISTER_FOR_NOTIFICATION:
/* Add this socket to the notify socket list*/
g_cluster.notify_clients = lappend_int(g_cluster.notify_clients, IPCCommand->issueing_sock);
@@ -1576,7 +1548,7 @@ static bool fire_node_status_event(int nodeID, int nodeStatus)
if (nodeStatus == WD_LIFECHECK_NODE_STATUS_DEAD)
{
ereport(DEBUG1,
- (errmsg("Firing NODE STATUS EVENT: NODE(ID=%d) IS DEAD",nodeID)));
+ (errmsg("firing NODE STATUS EVENT: NODE(ID=%d) IS DEAD",nodeID)));
if (wdNode == g_cluster.localNode)
watchdog_state_machine(WD_EVENT_LOCAL_NODE_LOST, wdNode, NULL);
@@ -1586,7 +1558,7 @@ static bool fire_node_status_event(int nodeID, int nodeStatus)
else if (nodeStatus == WD_LIFECHECK_NODE_STATUS_ALIVE)
{
ereport(DEBUG1,
- (errmsg("Firing NODE STATUS EVENT: NODE(ID=%d) IS ALIVE",nodeID)));
+ (errmsg("firing NODE STATUS EVENT: NODE(ID=%d) IS ALIVE",nodeID)));
if (wdNode == g_cluster.localNode)
watchdog_state_machine(WD_EVENT_LOCAL_NODE_FOUND, wdNode, NULL);
@@ -1644,169 +1616,6 @@ static IPC_CMD_PREOCESS_RES process_IPC_replicate_variable(WDIPCCommandData* IPC
return IPC_CMD_ERROR;
}
-static IPC_CMD_PREOCESS_RES process_IPC_unlock_request(WDIPCCommandData *IPCCommand)
-{
- IPC_CMD_PREOCESS_RES ret = IPC_CMD_COMPLETE;
- char res_type = WD_IPC_CMD_RESULT_OK;
- /*
- * if cluster or myself is not in stable state
- * just return cluster in transaction
- */
- IPCCommand->type = WD_INTERUNLOCKING_REQUEST;
-
- if (g_cluster.lockHolderNode == NULL)
- {
- /* There is no lock holder as per our records
- * just ignore this request
- */
- res_type = WD_IPC_CMD_RESULT_OK;
- }
- else if(g_cluster.lockHolderNode != g_cluster.localNode)
- {
- /* I am not the lockhoder node, so How can I unlock
- * just return the error
- */
- ret = WD_IPC_CMD_RESULT_BAD;
- }
- else if (get_local_node_state() == WD_STANDBY)
- {
- /* I am a standby node, and also the lock holder
- * Just forward the request to coordinator */
-
- WDPacketData *wdPacket = get_minimum_message(WD_INTERUNLOCKING_REQUEST,NULL);
- /* save the command ID */
- IPCCommand->internal_command_id = wdPacket->command_id;
-
- if (send_message(g_cluster.masterNode, wdPacket) <= 0)
- {
- /* we have failed to send to master node, return lock failed */
- res_type = WD_IPC_CMD_RESULT_BAD;
- }
- else
- {
- /*
- * we need to wait for the results from master/coordinator node
- * do not reply back to ipc until we hear from master node
- */
-
- ret = IPC_CMD_PROCESSING;
- }
- /* whatever the case we need to resign ourself from lock
- * holder
- */
- pfree(wdPacket);
- g_cluster.lockHolderNode = NULL;
- }
- else if (get_local_node_state() == WD_COORDINATOR)
- {
- /*
- * If I am coordinator, Just process the request locally
- */
- if (node_has_resigned_from_interlocking(g_cluster.localNode, NULL))
- {
- res_type = WD_IPC_CMD_RESULT_OK;
- }
- else
- {
- res_type = WD_IPC_CMD_RESULT_BAD;
- }
- }
- else /* we are not in any stable state at the moment */
- {
- g_cluster.lockHolderNode = NULL;
- res_type = WD_IPC_CMD_CLUSTER_IN_TRAN;
- }
-
- if (ret == IPC_CMD_PROCESSING)
- {
- /* Do not reply back to requester, as we are
- * still processing the results
- */
- return ret;
- }
- if (write_ipc_command_with_result_data(IPCCommand, res_type, NULL, 0))
- {
- /*
- * This is the complete lifecycle of command.
- * we are done with it
- */
- ret = IPC_CMD_COMPLETE;
- }
- else
- ret = IPC_CMD_ERROR;
-
- return ret;
-}
-
-
-static IPC_CMD_PREOCESS_RES process_IPC_lock_request(WDIPCCommandData *IPCCommand)
-{
- char res_type = WD_IPC_CMD_RESULT_BAD;
- /*
- * if cluster or myself is not in stable state
- * just return cluster in transaction
- */
- ereport(LOG,
- (errmsg("processing lock request from IPC socket")));
-
- IPCCommand->type = WD_INTERLOCKING_REQUEST;
-
- if (get_local_node_state() == WD_STANDBY)
- {
-
- /* I am a standby node, Just forward the request to coordinator */
- WDPacketData * wdPacket = get_minimum_message(WD_INTERLOCKING_REQUEST,NULL);
- /* save the command ID */
- IPCCommand->internal_command_id = wdPacket->command_id;
-
- if (send_message(g_cluster.masterNode, wdPacket) <= 0)
- {
- ereport(LOG,
- (errmsg("failed to process lock request from IPC socket"),
- errdetail("failed to forward the request to master watchdog node \"%s\"",g_cluster.masterNode->nodeName)));
-
- /* we have failed to send to any node, return lock failed */
- res_type = WD_IPC_CMD_RESULT_BAD;
- }
- else
- {
- /*
- * we need to wait for the result
- */
- ereport(LOG,
- (errmsg("lock request from IPC socket is forwarded to master watchdog node \"%s\"",g_cluster.masterNode->nodeName),
- errdetail("waiting for the reply from master node...")));
-
- return IPC_CMD_PROCESSING;
- }
- }
- else if (get_local_node_state() == WD_COORDINATOR)
- {
- /*
- * If I am coordinator, Just process the request locally
- */
- if (node_has_requested_for_interlocking(g_cluster.localNode, NULL))
- {
- res_type = WD_IPC_CMD_RESULT_OK;
- }
- else
- res_type = WD_IPC_CMD_RESULT_BAD;
- }
- else /* we are not in any stable state at the moment */
- {
- res_type = WD_IPC_CMD_CLUSTER_IN_TRAN;
- }
-
- if (write_ipc_command_with_result_data(IPCCommand, res_type, NULL, 0))
- {
- /*
- * This is the complete lifecycle of command.
- * we are done with it
- */
- return IPC_CMD_COMPLETE;
- }
- return IPC_CMD_ERROR;
-}
static IPC_CMD_PREOCESS_RES process_IPC_failover_cmd_synchronise(WDIPCCommandData *IPCCommand)
{
@@ -2315,6 +2124,9 @@ static void wd_system_will_go_down(int code, Datum arg)
close_socket_connection(&wdNode->client_socket);
close_socket_connection(&wdNode->server_socket);
}
+ /* close network monitoring socket */
+ if (g_cluster.network_monitor_sock > 0)
+ close(g_cluster.network_monitor_sock);
}
static void close_socket_connection(SocketConnection* conn)
@@ -3180,24 +2992,6 @@ static int update_connected_node_count(void)
return g_cluster.aliveNodeCount;
}
-static void update_nodes_connection_status(void)
-{
- int i;
- for (i = 0; i< g_cluster.remoteNodeCount; i++)
- {
- bool conectable = false;
- WatchdogNode* wdNode = &(g_cluster.remoteNodes[i]);
- if (is_socket_connection_connected(&wdNode->client_socket))
- conectable = true;
- else if (is_socket_connection_connected(&wdNode->server_socket))
- conectable = true;
- if (wdNode->is_connectable && conectable) /* new and old status is same */
- continue;
- wdNode->is_connectable = conectable;
- watchdog_state_machine(wdNode->is_connectable?WD_EVENT_NODE_CON_FOUND:WD_EVENT_NODE_CON_LOST, wdNode, NULL);
- }
-}
-
static bool service_lost_connections(void)
{
@@ -3448,7 +3242,7 @@ static void cluster_in_stable_state(void)
static int watchdog_state_machine(WD_EVENTS event, WatchdogNode* wdNode, WDPacketData* pkt)
{
ereport(DEBUG1,
- (errmsg("STATE MACHINE INVOKED WITH EVENT = %s Current State = %s",wd_event_name[event], debug_states[get_local_node_state()])));
+ (errmsg("STATE MACHINE INVOKED WITH EVENT = %s Current State = %s",wd_event_name[event], wd_state_names[get_local_node_state()])));
if (event == WD_EVENT_REMOTE_NODE_LOST)
{
@@ -4554,7 +4348,7 @@ static int set_state(WD_STATES newState)
if (oldState != newState)
{
ereport(LOG,
- (errmsg("watchdog node state changed from [%s] to [%s]",debug_states[oldState],debug_states[newState])));
+ (errmsg("watchdog node state changed from [%s] to [%s]",wd_state_names[oldState],wd_state_names[newState])));
watchdog_state_machine(WD_EVENT_WD_STATE_CHANGED, NULL, NULL);
/* send out the info message to all nodes */
send_message_of_type(NULL, WD_INFO_MESSAGE);
@@ -5102,7 +4896,7 @@ static void print_watchdog_node_info(WatchdogNode* wdNode)
{
ereport(DEBUG2,
(errmsg("state: \"%s\" Host: \"%s\" Name: \"%s\" WD Port:%d PP Port: %d priority:%d",
- debug_states[wdNode->state],
+ wd_state_names[wdNode->state],
wdNode->hostname
,wdNode->nodeName
,wdNode->wd_port
@@ -5129,5 +4923,5 @@ static void print_received_packet_info(WDPacketData* pkt,WatchdogNode* wdNode)
(errmsg("watchdog packet received from node \"%s\"",wdNode->nodeName),
errdetail("command id : %d Type: %s my watchdog state :%s",pkt->command_id,
pkt_type?pkt_type->name:"UNKNOWN",
- debug_states[get_local_node_state()])));
+ wd_state_names[get_local_node_state()])));
}
diff --git a/src/watchdog/wd_commands.c b/src/watchdog/wd_commands.c
index 0097afa01..87c1ba4d0 100644
--- a/src/watchdog/wd_commands.c
+++ b/src/watchdog/wd_commands.c
@@ -20,7 +20,6 @@
* is" without express or implied warranty.
*
*/
-#include <pthread.h>
#include <stdio.h>
#include <errno.h>
#include <ctype.h>
@@ -50,6 +49,7 @@
#include "watchdog/wd_ipc_commands.h"
#include "watchdog/wd_ipc_defines.h"
+#define WD_DEFAULT_IPC_COMMAND_TIMEOUT 5 /* default number of seconds to wait for IPC command results*/
#define WD_INTERLOCK_WAIT_MSEC 500
#define WD_INTERLOCK_TIMEOUT_SEC 10
#define WD_INTERLOCK_WAIT_COUNT ((int) ((WD_INTERLOCK_TIMEOUT_SEC * 1000)/WD_INTERLOCK_WAIT_MSEC))
@@ -93,7 +93,7 @@ void wd_ipc_initialize_data(void)
pool_config->wd_ipc_socket_dir,
pool_config->wd_port);
- watchdog_ipc_address = pool_shared_memory_create(strlen(wd_ipc_sock_addr));
+ watchdog_ipc_address = pool_shared_memory_create(strlen(wd_ipc_sock_addr) +1);
strcpy(watchdog_ipc_address, wd_ipc_sock_addr);
}
}
@@ -103,38 +103,42 @@ char* get_watchdog_ipc_address(void)
return watchdog_ipc_address;
}
+/*
+ * function issues the command to watchdog process over the watchdog
+ * IPC command socket.
+ * type: command type to send. valid command
+ * types are defined in wd_ipc_defines.h
+ * timeout_sec: number of seconds to wait for the command response
+ * from watchdog
+ * data: command data
+ * data_len: length of data
+ * blocking: send true if caller wants to wait for the results
+ * when blocking is false the timeout_sec is ignored
+ */
WDIPCCmdResult*
-issue_command_to_watchdog(char type, WD_COMMAND_ACTIONS command_action,int timeout_sec, char* data, int data_len, bool blocking)
+issue_command_to_watchdog(char type, int timeout_sec, char* data, int data_len, bool blocking)
{
struct timeval start_time,tv;
int sock;
WDIPCCmdResult* result = NULL;
char res_type = 'P';
- int res_length, action;
+ int res_length, len;
gettimeofday(&start_time, NULL);
-
+
/* open the watchdog command socket for IPC */
sock = open_wd_command_sock(false);
if (sock < 0)
return NULL;
- res_length = htonl(data_len);
- action = htonl(command_action);
+ len = htonl(data_len);
if (socket_write(sock, &type, sizeof(char)) <= 0)
{
close(sock);
return NULL;
}
-
-
- if (socket_write(sock, &action, sizeof(int)) <= 0)
- {
- close(sock);
- return NULL;
- }
- if (socket_write(sock, &res_length, sizeof(int)) <= 0)
+ if (socket_write(sock, &len, sizeof(int)) <= 0)
{
close(sock);
return NULL;
@@ -169,7 +173,7 @@ issue_command_to_watchdog(char type, WD_COMMAND_ACTIONS command_action,int timeo
/* read the result type char */
if (socket_read(sock, &res_type, 1 ,0) <=0)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("error reading from IPC command socket"),
errdetail("read from socket failed with error \"%s\"",strerror(errno))));
close(sock);
@@ -178,7 +182,7 @@ issue_command_to_watchdog(char type, WD_COMMAND_ACTIONS command_action,int timeo
/* read the result data length */
if (socket_read(sock, &res_length, sizeof(int), 0) <= 0)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("error reading from IPC command socket"),
errdetail("read from socket failed with error \"%s\"",strerror(errno))));
close(sock);
@@ -227,12 +231,14 @@ wd_start_recovery(void)
char type;
char* func = get_wd_node_function_json(WD_FUNCTION_START_RECOVERY, NULL,0);
- WDIPCCmdResult *result = issue_command_to_watchdog(WD_FUNCTION_COMMAND, WD_COMMAND_ACTION_DEFAULT,pool_config->recovery_timeout, func, strlen(func), true);
+ WDIPCCmdResult *result = issue_command_to_watchdog(WD_FUNCTION_COMMAND,
+ pool_config->recovery_timeout,
+ func, strlen(func), true);
pfree(func);
if (result == NULL)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("start recovery command lock failed"),
errdetail("issue command to watchdog returned NULL")));
return COMMAND_FAILED;
@@ -242,7 +248,7 @@ wd_start_recovery(void)
pfree(result);
if (type == WD_IPC_CMD_CLUSTER_IN_TRAN)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("start recovery command lock failed"),
errdetail("watchdog cluster is not in stable state"),
errhint("try again when the cluster is fully initialized")));
@@ -260,12 +266,14 @@ wd_end_recovery(void)
char type;
char* func = get_wd_node_function_json(WD_FUNCTION_END_RECOVERY, NULL, 0);
- WDIPCCmdResult *result = issue_command_to_watchdog(WD_FUNCTION_COMMAND, WD_COMMAND_ACTION_DEFAULT,2, func, strlen(func), true);
+ WDIPCCmdResult *result = issue_command_to_watchdog(WD_FUNCTION_COMMAND,
+ WD_DEFAULT_IPC_COMMAND_TIMEOUT,
+ func, strlen(func), true);
pfree(func);
if (result == NULL)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("start recovery command lock failed"),
errdetail("issue command to watchdog returned NULL")));
return COMMAND_FAILED;
@@ -275,7 +283,7 @@ wd_end_recovery(void)
pfree(result);
if (type == WD_IPC_CMD_CLUSTER_IN_TRAN)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("start recovery command lock failed"),
errdetail("watchdog cluster is not in stable state"),
errhint("try again when the cluster is fully initialized")));
@@ -301,12 +309,14 @@ wd_send_failback_request(int node_id)
return COMMAND_OK;
func = get_wd_node_function_json(WD_FUNCTION_FAILBACK_REQUEST,&n, 1);
- WDIPCCmdResult *result = issue_command_to_watchdog(WD_FUNCTION_COMMAND, WD_COMMAND_ACTION_DEFAULT,2, func, strlen(func), true);
+ WDIPCCmdResult *result = issue_command_to_watchdog(WD_FUNCTION_COMMAND,
+ WD_DEFAULT_IPC_COMMAND_TIMEOUT,
+ func, strlen(func), true);
pfree(func);
if (result == NULL)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("start recovery command lock failed"),
errdetail("issue command to watchdog returned NULL")));
return COMMAND_FAILED;
@@ -316,7 +326,7 @@ wd_send_failback_request(int node_id)
pfree(result);
if (type == WD_IPC_CMD_CLUSTER_IN_TRAN)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("start recovery command lock failed"),
errdetail("watchdog cluster is not in stable state"),
errhint("try again when the cluster is fully initialized")));
@@ -351,13 +361,15 @@ wd_send_failover_sync_command(WDFailoverCMDTypes cmdType, char* syncReqType)
char* json_data = get_wd_failover_cmd_type_json(cmdType, syncReqType);
- WDIPCCmdResult *result = issue_command_to_watchdog(WD_FAILOVER_CMD_SYNC_REQUEST, WD_COMMAND_ACTION_DEFAULT,pool_config->recovery_timeout, json_data, strlen(json_data), true);
+ WDIPCCmdResult *result = issue_command_to_watchdog(WD_FAILOVER_CMD_SYNC_REQUEST
+ ,pool_config->recovery_timeout,
+ json_data, strlen(json_data), true);
pfree(json_data);
if (result == NULL || result->length <= 0)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("start recovery command lock failed"),
errdetail("issue command to watchdog returned NULL")));
return FAILOVER_RES_ERROR;
@@ -393,44 +405,6 @@ wd_send_failover_sync_command(WDFailoverCMDTypes cmdType, char* syncReqType)
return interlockingResult;
}
-WdCommandResult
-wd_try_command_lock(void)
-{
- WDIPCCmdResult* result;
- char type;
-
- result = issue_command_to_watchdog(WD_TRY_COMMAND_LOCK, WD_COMMAND_ACTION_DEFAULT,10, NULL, 0, true);
- if (result == NULL)
- {
- ereport(LOG,
- (errmsg("watchdog command lock failed"),
- errdetail("issue command to watchdog returned NULL")));
- return COMMAND_FAILED;
- }
-
- type = result->type;
- pfree(result);
- if (type == WD_IPC_CMD_CLUSTER_IN_TRAN)
- {
- ereport(LOG,
- (errmsg("watchdog command lock failed"),
- errdetail("watchdog cluster is not in stable state"),
- errhint("try again when the cluster is fully initialized")));
- return CLUSTER_IN_TRANSATIONING;
- }
-
- if (type == WD_IPC_CMD_RESULT_OK)
- return COMMAND_OK;
-
- return COMMAND_FAILED;
-}
-
-void wd_command_unlock(void)
-{
- /* we dont really care about results here */
- issue_command_to_watchdog(WD_COMMAND_UNLOCK, WD_COMMAND_ACTION_DEFAULT,10, NULL, 0, false);
-}
-
WdCommandResult
wd_degenerate_backend_set(int *node_id_set, int count)
{
@@ -443,13 +417,15 @@ wd_degenerate_backend_set(int *node_id_set, int count)
return COMMAND_OK;
func = get_wd_node_function_json(WD_FUNCTION_DEGENERATE_REQUEST,node_id_set, count);
- WDIPCCmdResult *result = issue_command_to_watchdog(WD_FUNCTION_COMMAND, WD_COMMAND_ACTION_DEFAULT,2, func, strlen(func), true);
+ WDIPCCmdResult *result = issue_command_to_watchdog(WD_FUNCTION_COMMAND ,
+ WD_DEFAULT_IPC_COMMAND_TIMEOUT,
+ func, strlen(func), true);
pfree(func);
if (result == NULL)
{
- ereport(LOG,
- (errmsg("degenerate backend set command failed"),
+ ereport(WARNING,
+ (errmsg("degenerate backend set command failed"),
errdetail("issue command to watchdog returned NULL")));
return COMMAND_FAILED;
}
@@ -458,7 +434,7 @@ wd_degenerate_backend_set(int *node_id_set, int count)
pfree(result);
if (type == WD_IPC_CMD_CLUSTER_IN_TRAN)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("degenerate backend set command failed"),
errdetail("watchdog cluster is not in stable state"),
errhint("try again when the cluster is fully initialized")));
@@ -483,12 +459,14 @@ wd_promote_backend(int node_id)
return COMMAND_OK;
func = get_wd_node_function_json(WD_FUNCTION_PROMOTE_REQUEST,&n, 1);
- result = issue_command_to_watchdog(WD_FUNCTION_COMMAND, WD_COMMAND_ACTION_DEFAULT,2, func, strlen(func), true);
+ result = issue_command_to_watchdog(WD_FUNCTION_COMMAND ,
+ WD_DEFAULT_IPC_COMMAND_TIMEOUT,
+ func, strlen(func), true);
pfree(func);
if (result == NULL)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("start recovery command lock failed"),
errdetail("issue command to watchdog returned NULL")));
return COMMAND_FAILED;
@@ -498,7 +476,7 @@ wd_promote_backend(int node_id)
pfree(result);
if (type == WD_IPC_CMD_CLUSTER_IN_TRAN)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("start recovery command lock failed"),
errdetail("watchdog cluster is not in stable state"),
errhint("try again when the cluster is fully initialized")));
@@ -524,13 +502,15 @@ char* wd_get_watchdog_nodes(int nodeID)
json_str = jw_get_json_string(jNode);
- result = issue_command_to_watchdog(WD_GET_NODES_LIST_COMMAND, WD_COMMAND_ACTION_DEFAULT,5, json_str, strlen(json_str), true);
+ result = issue_command_to_watchdog(WD_GET_NODES_LIST_COMMAND
+ ,WD_DEFAULT_IPC_COMMAND_TIMEOUT,
+ json_str, strlen(json_str), true);
jw_destroy(jNode);
if (result == NULL)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("get watchdog nodes command failed"),
errdetail("issue command to watchdog returned NULL")));
return NULL;
@@ -538,7 +518,7 @@ char* wd_get_watchdog_nodes(int nodeID)
if (result->type == WD_IPC_CMD_CLUSTER_IN_TRAN)
{
- ereport(LOG,
+ ereport(WARNING,
(errmsg("get watchdog nodes command failed"),
errdetail("watchdog cluster is not in stable state"),
errhint("try again when the cluster is fully initialized")));
diff --git a/src/watchdog/wd_escalation.c b/src/watchdog/wd_escalation.c
index 5c97a09fb..16349323a 100644
--- a/src/watchdog/wd_escalation.c
+++ b/src/watchdog/wd_escalation.c
@@ -35,7 +35,6 @@
#include "watchdog/wd_utils.h"
#include "query_cache/pool_memqcache.h"
-extern int *escalation_status;
static void
wd_exit(int exit_signo)
@@ -59,8 +58,6 @@ pid_t
fork_escalation_process(void)
{
pid_t pid;
- int has_vip = 0;
-
pid = fork();
if (pid != 0)
{
@@ -85,13 +82,10 @@ fork_escalation_process(void)
MemoryContextSwitchTo(TopMemoryContext);
- set_ps_display("escalation",false);
+ set_ps_display("watchdog escalation",false);
ereport(LOG,
(errmsg("watchdog: escalation started")));
-
- *escalation_status = 1;
-
/*
* STEP 1
* clear shared memory cache
@@ -108,7 +102,7 @@ fork_escalation_process(void)
/*
* STEP 2
- * execute escalation command
+ * execute escalation command provided by user in pgpool conf file
*/
if (strlen(pool_config->wd_escalation_command))
{
@@ -133,17 +127,15 @@ fork_escalation_process(void)
/*
* STEP 3
- * interface up as delegate IP
+ * bring up the delegate IP
*/
-
if (strlen(pool_config->delegate_IP) != 0)
{
- has_vip = wd_IP_up();
- }
-
- if (has_vip == WD_OK)
- *escalation_status = 2;
+ if (wd_IP_up() != WD_OK)
+ ereport(WARNING,
+ (errmsg("watchdog de-escalation failed to bring down delegate IP")));
+ }
exit(0);
}
@@ -154,8 +146,7 @@ pid_t
fork_plunging_process(void)
{
pid_t pid;
- int has_vip = 0;
-
+
pid = fork();
if (pid != 0)
{
@@ -180,16 +171,14 @@ fork_plunging_process(void)
MemoryContextSwitchTo(TopMemoryContext);
- set_ps_display("de-escalation",false);
+ set_ps_display("watchdog de-escalation",false);
ereport(LOG,
(errmsg("watchdog: de-escalation started")));
- *escalation_status = 1;
-
/*
- * STEP 2
- * execute escalation command
+ * STEP 1
+ * execute de-escalation command provided by user in pgpool conf file
*/
if (strlen(pool_config->wd_de_escalation_command))
{
@@ -213,16 +202,15 @@ fork_plunging_process(void)
}
/*
- * STEP 3
- * interface up as delegate IP
+ * STEP 2
+ * bring down the delegate IP
*/
if (strlen(pool_config->delegate_IP) != 0)
{
- has_vip = wd_IP_down();
+ if (wd_IP_down() != WD_OK)
+ ereport(WARNING,
+ (errmsg("watchdog de-escalation failed to bring down delegate IP")));
}
-
- if (has_vip == WD_OK)
- *escalation_status = 2;
exit(0);
}
diff --git a/src/watchdog/wd_interlock.c b/src/watchdog/wd_interlock.c
deleted file mode 100644
index 77534b3b1..000000000
--- a/src/watchdog/wd_interlock.c
+++ /dev/null
@@ -1,150 +0,0 @@
-/*
- * $Header$
- *
- * Handles watchdog connection, and protocol communication with pgpool-II
- *
- * pgpool: a language independent connection pool server for PostgreSQL
- * written by Tatsuo Ishii
- *
- * Copyright (c) 2003-2013 PgPool Global Development Group
- *
- * Permission to use, copy, modify, and distribute this software and
- * its documentation for any purpose and without fee is hereby
- * granted, provided that the above copyright notice appear in all
- * copies and that both that copyright notice and this permission
- * notice appear in supporting documentation, and that the name of the
- * author not be used in advertising or publicity pertaining to
- * distribution of the software without specific, written prior
- * permission. The author makes no representations about the
- * suitability of this software for any purpose. It is provided "as
- * is" without express or implied warranty.
- *
- */
-
-/*
- interlocking mechnism
-
- 1-) Only master can assign locks
- 2-) pgpool-II Will ask watchdog to aquire lock
- -- watchdog will return one of following
- -- CLUSTER IN TRANSATIONING
- -- LOCK AQUIRED
- -- LOCK FAILED
-
- 3-) On receiving lock request, standby pgpool-II will forward
- the lock request to master node. And wait for the reply
- and return the result to caller based on reply
-
- 4-) on receiving lock request, master pgpool-II will replicate
- check if lock holder exists, it will decline the request
- and accept it otherwise.
- -- Once the new lock is granted this information will be
- replicated to all connected nodes.
-
- */
-
-#include <stdio.h>
-#include <string.h>
-#include <stdlib.h>
-#include <sys/time.h>
-#include <unistd.h>
-
-#include "pool.h"
-#include "utils/elog.h"
-#include "pool_config.h"
-#include "watchdog/watchdog.h"
-#include "watchdog/wd_ext.h"
-
-static volatile bool * WD_Locks;
-
-static void wd_update_lock_holder(void);
-static int wd_assume_lock_holder(void);
-static void wd_resign_lock_holder(void);
-static void wd_lock_all(void);
-static void wd_confirm_contactable(void);
-
-static void sleep_in_waiting(void);
-static WDFailoverCMDResults wd_issue_failover_lock_command(WDFailoverCMDTypes cmdType, char* syncReqType);
-
-#define WD_INTERLOCK_WAIT_MSEC 500
-#define WD_INTERLOCK_TIMEOUT_SEC 10
-#define WD_INTERLOCK_WAIT_COUNT ((int) ((WD_INTERLOCK_TIMEOUT_SEC * 1000)/WD_INTERLOCK_WAIT_MSEC))
-
-
-void wd_wati_until_lock_or_timeout(WDFailoverCMDTypes cmdType)
-{
- WDFailoverCMDResults res = FAILOVER_RES_TRANSITION;
- int count = WD_INTERLOCK_WAIT_COUNT;
- while (1)
- {
- res = wd_failover_command_check_lock(cmdType);
- if (res == FAILOVER_RES_PROCEED_LOCK_HOLDER ||
- res == FAILOVER_RES_PROCEED_UNLOCKED)
- {
- /* we have the permision */
- return;
- }
- sleep_in_waiting();
- if (--count < 0)
- {
- ereport(WARNING,
- (errmsg("timeout wating for unlock")));
- break;
- }
- }
-}
-
-WDFailoverCMDResults wd_failover_command_start(WDFailoverCMDTypes cmdType)
-{
- if (pool_config->use_watchdog)
- return wd_issue_failover_lock_command(cmdType,"START_COMMAND");
- return FAILOVER_RES_PROCEED_LOCK_HOLDER;
-}
-
-WDFailoverCMDResults wd_failover_command_end(WDFailoverCMDTypes cmdType)
-{
- if (pool_config->use_watchdog)
- return wd_issue_failover_lock_command(cmdType,"END_COMMAND");
- return FAILOVER_RES_PROCEED_LOCK_HOLDER;
-}
-
-WDFailoverCMDResults wd_failover_command_check_lock(WDFailoverCMDTypes cmdType)
-{
- if (pool_config->use_watchdog)
- return wd_issue_failover_lock_command(cmdType,"CHECK_LOCKED");
- return FAILOVER_RES_PROCEED_LOCK_HOLDER;
-}
-
-WDFailoverCMDResults wd_release_failover_command_lock(WDFailoverCMDTypes cmdType)
-{
- if (pool_config->use_watchdog)
- return wd_issue_failover_lock_command(cmdType,"UNLOCK_COMMAND");
- return FAILOVER_RES_PROCEED_LOCK_HOLDER;
-}
-
-/*
- * This is just a wrapper over wd_send_failover_sync_command()
- * but tries to wait for WD_INTERLOCK_TIMEOUT_SEC amount of time
- * if watchdog is in transition state
- */
-
-static WDFailoverCMDResults wd_issue_failover_lock_command(WDFailoverCMDTypes cmdType, char* syncReqType)
-{
- WDFailoverCMDResults res;
- int x;
- for (x=0; x < MAX_SEC_WAIT_FOR_CLUSTER_TRANSATION; x++)
- {
- res = wd_send_failover_sync_command(NODE_FAILBACK_CMD, syncReqType);
- if (res != FAILOVER_RES_TRANSITION)
- break;
- sleep(1);
- }
- return res;
-}
-
-static void
-sleep_in_waiting(void)
-{
- struct timeval t = {0, WD_INTERLOCK_WAIT_MSEC * 1000};
- select(0, NULL, NULL, NULL, &t);
-}
diff --git a/src/watchdog/wd_lifecheck.c b/src/watchdog/wd_lifecheck.c
index 7652e3bae..8d638a73b 100644
--- a/src/watchdog/wd_lifecheck.c
+++ b/src/watchdog/wd_lifecheck.c
@@ -48,6 +48,12 @@
#include "libpq-fe.h"
+#define LIFECHECK_GETNODE_WAIT_SEC_COUNT 5 /* max number of seconds the lifecheck process
+ * should waits before giving up
+ * while fetching the configured watchdog node
+ * infromation from watchdog process through IPC channel
+ */
+
/*
* thread argument for lifecheck of pgpool
*/
@@ -312,6 +318,7 @@ static pid_t fork_lifecheck_child(void)
return pid;
}
+
/* main entry point of watchdog lifecheck process*/
static pid_t
lifecheck_main(void)
@@ -346,11 +353,11 @@ lifecheck_main(void)
/* Get the list of watchdog node to monitor
* from watchdog process
*/
- for (i =0; i < 5; i++)
+ for (i =0; i < LIFECHECK_GETNODE_WAIT_SEC_COUNT; i++)
{
if (fetch_watchdog_nodes_data() == true)
break;
- sleep(2);
+ sleep(1);
}
if (!gslifeCheckCluster)
@@ -466,7 +473,7 @@ static bool inform_node_status(LifeCheckNode* node, char *message)
for (x=0; x < MAX_SEC_WAIT_FOR_CLUSTER_TRANSATION; x++)
{
- res = issue_command_to_watchdog(WD_NODE_STATUS_CHANGE_COMMAND, WD_COMMAND_ACTION_DEFAULT,0, json_data, strlen(json_data),false);
+ res = issue_command_to_watchdog(WD_NODE_STATUS_CHANGE_COMMAND ,0, json_data, strlen(json_data),false);
if (res)
break;
sleep(1);
@@ -630,7 +637,7 @@ static int wd_lifecheck(void)
ereport(WARNING,
(errmsg("watchdog lifecheck, failed to connect to any trusted servers")));
- inform_node_status(node,"trusted server is down");
+ inform_node_status(node,"trusted server is unreachable");
}
return WD_NG;
}
diff --git a/src/watchdog/wd_list.c b/src/watchdog/wd_list.c
deleted file mode 100644
index 6179a1db1..000000000
--- a/src/watchdog/wd_list.c
+++ /dev/null
@@ -1,403 +0,0 @@
-/*
- * $Header$
- *
- * Handles watchdog connection, and protocol communication with pgpool-II
- *
- * pgpool: a language independent connection pool server for PostgreSQL
- * written by Tatsuo Ishii
- *
- * Copyright (c) 2003-2012 PgPool Global Development Group
- *
- * Permission to use, copy, modify, and distribute this software and
- * its documentation for any purpose and without fee is hereby
- * granted, provided that the above copyright notice appear in all
- * copies and that both that copyright notice and this permission
- * notice appear in supporting documentation, and that the name of the
- * author not be used in advertising or publicity pertaining to
- * distribution of the software without specific, written prior
- * permission. The author makes no representations about the
- * suitability of this software for any purpose. It is provided "as
- * is" without express or implied warranty.
- *
- */
-
-#include <stdio.h>
-#include <string.h>
-#include <stdlib.h>
-#include <sys/time.h>
-
-#include "pool.h"
-#include "utils/elog.h"
-#include "pool_config.h"
-#include "watchdog/watchdog.h"
-#include "watchdog/wd_ext.h"
-
-static void wd_reset_wd_info(WdInfo *info);
-
-/* add or modify watchdog information list */
-int
-wd_set_wd_list(char * hostname, int pgpool_port, int wd_port, char * delegate_ip, struct timeval * tv, int status)
-{
- int i = 0;
- WdInfo * p = NULL;
-
- if ((WD_List == NULL) || (hostname == NULL))
- ereport(ERROR,
- (errmsg("adding watchdog information list. memory allocation error")));
-
- if (strcmp(pool_config->delegate_IP, delegate_ip))
- ereport(ERROR,
- (errmsg("adding watchdog information list. delegate IP mismatch error"),
- errdetail("delegate IP defined in config \"%s\" does not match with \"%s\"",pool_config->delegate_IP, delegate_ip)));
-
- for ( i = 0 ; i < MAX_WATCHDOG_NUM ; i ++)
- {
- p = (WD_List+i);
-
- if( p->status != WD_END)
- {
- /* found; modify the pgpool. */
- if ((!strncmp(p->hostname, hostname, sizeof(p->hostname))) &&
- (p->pgpool_port == pgpool_port) &&
- (p->wd_port == wd_port))
- {
- if (p->status == WD_DOWN || p->status == WD_INIT)
- {
- wd_reset_wd_info(p);
- }
-
- p->status = status;
-
- if (tv != NULL)
- {
- memcpy(&(p->tv), tv, sizeof(struct timeval));
- }
-
- return i;
- }
- }
-
- /* not found; add as a new pgpool */
- else
- {
- wd_reset_wd_info(p);
-
- p->status = status;
- p->pgpool_port = pgpool_port;
- p->wd_port = wd_port;
-
- strlcpy(p->hostname, hostname, sizeof(p->hostname));
- strlcpy(p->delegate_ip, delegate_ip, sizeof(p->delegate_ip));
-
- if (tv != NULL)
- {
- memcpy(&(p->tv), tv, sizeof(struct timeval));
- }
-
- return i;
- }
- }
- ereport(WARNING,
- (errmsg("failed adding watchdog information list. list is full")));
- return -1;
-}
-
-/** reset watchdog info*/
-static void
-wd_reset_wd_info(WdInfo * info)
-{
- WD_TIME_INIT(info->hb_send_time);
- WD_TIME_INIT(info->hb_last_recv_time);
-
- info->in_interlocking = false;
- info->is_lock_holder = false;
-}
-
-/* add watchdog information to list using config description */
-int
-wd_add_wd_list(WdDesc * other_wd)
-{
- WdInfo * p = NULL;
- int i = 0;
-
- if (other_wd == NULL)
- {
- ereport(ERROR,
- (errmsg("adding watchdog information list. memory allocation error")));
- }
- for ( i = 0 ; i < other_wd->num_wd ; i ++)
- {
- p = &(other_wd->wd_info[i]);
- strlcpy(p->delegate_ip, pool_config->delegate_IP, sizeof(p->delegate_ip));
- wd_set_wd_info(p);
- }
-
- return i;
-}
-
-/* set watchdog information to list */
-int
-wd_set_wd_info(WdInfo * info)
-{
- int rtn;
- rtn = wd_set_wd_list(info->hostname, info->pgpool_port, info->wd_port,
- info->delegate_ip, &(info->tv), info->status);
- return rtn;
-}
-
-/* return master if exist, NULL if not found */
-WdInfo *
-wd_is_exist_master(void)
-{
- WdInfo * p = WD_List;
-
- while (p->status != WD_END)
- {
- /* find master pgpool in the other servers */
- if (p->status == WD_MASTER)
- {
- /* master found */
- return p;
- }
- p++;
- }
- /* not found */
- return NULL;
-}
-
-/* set or unset in_interlocking flag */
-void
-wd_set_interlocking(WdInfo *info, bool value)
-{
- WdInfo * p = WD_List;
-
- while (p->status != WD_END)
- {
- if ((!strncmp(p->hostname, info->hostname, sizeof(p->hostname))) &&
- (p->pgpool_port == info->pgpool_port) &&
- (p->wd_port == info->wd_port))
- {
- p->in_interlocking = value;
-
- return;
- }
- p++;
- }
-}
-
-/* set or unset lock holder flag */
-void
-wd_set_lock_holder(WdInfo *info, bool value)
-{
- WdInfo * p = WD_List;
-
- while (p->status != WD_END)
- {
- if ((!strncmp(p->hostname, info->hostname, sizeof(p->hostname))) &&
- (p->pgpool_port == info->pgpool_port) &&
- (p->wd_port == info->wd_port))
- {
- p->is_lock_holder = value;
-
- return;
- }
- p++;
- }
-}
-
-/* return the lock holder if exist, NULL if not found */
-WdInfo *
-wd_get_lock_holder(void)
-{
- WdInfo * p = WD_List;
-
- while (p->status != WD_END)
- {
- /* find failover lock holder */
- if ((p->status == WD_NORMAL || p->status == WD_MASTER) &&
- p->is_lock_holder)
- {
- /* found */
- return p;
- }
- p++;
- }
-
- /* not found */
- return NULL;
-}
-
-/* return the pgpool in interlocking found in first, NULL if not found */
-WdInfo *
-wd_get_interlocking(void)
-{
- WdInfo * p = WD_List;
-
- /* for updating contactable flag */
- wd_update_info();
-
- while (p->status != WD_END)
- {
- /* skip if not contactable */
- if (!p->is_contactable)
- {
- p++;
- continue;
- }
-
- /* find pgpool in interlocking */
- if ((p->status == WD_NORMAL || p->status == WD_MASTER) &&
- p->in_interlocking)
- {
- /* found */
- return p;
- }
- p++;
- }
-
- /* not found */
- return NULL;
-}
-
-/* if all pgpool are in interlocking return true, otherwise false */
-bool
-wd_are_interlocking_all(void)
-{
- WdInfo * p = WD_List;
- bool rtn = true;
-
- /* for updating contactable flag */
- wd_update_info();
-
- while (p->status != WD_END)
- {
- /* skip if not contactable */
- if (!p->is_contactable)
- {
- p++;
- continue;
- }
-
- /* find pgpool not in interlocking */
- if ((p->status == WD_NORMAL || p->status == WD_MASTER) &&
- !p->in_interlocking)
- {
- rtn = false;
- }
- p++;
- }
-
- return rtn;
-}
-
-/* clear flags for interlocking */
-void
-wd_clear_interlocking_info(void)
-{
- WdInfo * p = WD_List;
-
- while (p->status != WD_END)
- {
- wd_set_lock_holder(p, false);
- wd_set_interlocking(p, false);
- p++;
- }
-}
-
-int
-wd_am_I_oldest(void)
-{
- WdInfo * p = WD_List;
-
- p++;
- while (p->status != WD_END)
- {
- if ((p->status == WD_NORMAL) ||
- (p->status == WD_MASTER))
- {
- if (WD_TIME_BEFORE(p->tv, WD_MYSELF->tv))
- {
- return WD_NG;
- }
- }
- p++;
- }
- return WD_OK;
-}
-
-int
-wd_set_myself(struct timeval * tv, int status)
-{
- if (WD_MYSELF == NULL)
- {
- return WD_NG;
- }
-
- if (tv != NULL)
- {
- memcpy(&(WD_MYSELF->tv),tv,sizeof(struct timeval));
- }
-
- WD_MYSELF->status = status;
-
- return WD_OK;
-}
-
-
-
-/*
- * if master exists and it is contactable return true,
- * otherwise return false
- */
-bool
-wd_is_contactable_master(void)
-{
- WdInfo * master = NULL;
-
- if (WD_MYSELF->status == WD_MASTER)
- return true;
-
- /* for updating contactable flag */
- wd_update_info();
-
- master = wd_is_exist_master();
- if (master != NULL)
- {
- return master->is_contactable;
- }
-
- return false;
-}
-
-bool
-wd_are_contactable_all(void)
-{
- WdInfo * p = WD_List;
-
- /* for updating contactable flag */
- wd_update_info();
-
- while (p->status != WD_END)
- {
- if ((p->status == WD_NORMAL || p->status == WD_MASTER) &&
- !p->is_contactable)
- {
- return false;
- }
- p++;
- }
-
- return true;
-}
-
-/*
- * get watchdog information of specified index
- */
-WdInfo *
-wd_get_watchdog_info(int wd_index)
-{
- if (wd_index < 0 || wd_index > pool_config->other_wd->num_wd)
- return NULL;
-
- return &WD_List[wd_index];
-}
diff --git a/src/watchdog/wd_ping.c b/src/watchdog/wd_ping.c
index c5832eec4..6ad149020 100644
--- a/src/watchdog/wd_ping.c
+++ b/src/watchdog/wd_ping.c
@@ -33,7 +33,6 @@
#include "pool.h"
#include "utils/elog.h"
#include "pool_config.h"
-//#include "watchdog/watchdog.h"
#include "watchdog/wd_utils.h"
#define WD_MAX_PING_RESULT 256
[parent: 701890a7cb11]