From df5dd43e7c6a2c1cb71bf7ca1566c3a29a2fc1e3 Mon Sep 17 00:00:00 2001
From: Ronnie Sahlberg <sahlberg@ronnie>
Date: Mon, 5 Nov 2007 13:36:11 +1100
Subject: add a new tunable "CheckNodesFile" that when set to 0   will disable
 the check in the recovery daemon that all nodes are using the same
 /etc/ctdb/nodes file.

Also add some more missing checks that the pnn used is a valid pnn
before using it to dereferencing the ctdb->nodes array


This is useful since it allows us to add more physical nodes to a an
existing cluster without having to bring down the entire cluster.

The to add an additional node to an existing cluster would then be
1, on all nodes set CheckNodesFile=0 using 'ctdb setvar'
2, on all nodes add CTDB_SET_CheckNodesFile=0 to /etc/sysconfig/ctdb
For each each node, one at a time :
3, use 'ctdb disable' to stop the hosted services
4, service ctdb stop
5, service ctdb start
Once all nodes have been restarted
6, on all nodes remove CTDB_SET_CheckNodesFile=0 from
/etc/sysconfig/ctdb
7, on all nodes set CheckNodesFile=0 using 'ctdb setvar'

8, configure and start up the new node

During this procedure, only one node at a time was brought
down/restarted and was so only for a short period.

(This used to be ctdb commit 462501a32143e943ce350bd904a47c0955414a51)
---
 ctdb/include/ctdb_private.h   |  1 +
 ctdb/server/ctdb_persistent.c | 13 ++++++++++++-
 ctdb/server/ctdb_recoverd.c   | 26 +++++++++++++++++++++-----
 ctdb/server/ctdb_server.c     | 12 +++++++++++-
 ctdb/server/ctdb_takeover.c   |  6 ++++++
 ctdb/server/ctdb_tunables.c   |  1 +
 6 files changed, 52 insertions(+), 7 deletions(-)

(limited to 'ctdb')

diff --git a/ctdb/include/ctdb_private.h b/ctdb/include/ctdb_private.h
index 7b98683e9f..9df80b64af 100644
--- a/ctdb/include/ctdb_private.h
+++ b/ctdb/include/ctdb_private.h
@@ -88,6 +88,7 @@ struct ctdb_tunable {
 	uint32_t database_hash_size;
 	uint32_t rerecovery_timeout;
 	uint32_t enable_bans;
+	uint32_t check_nodes_file;
 	uint32_t deterministic_public_ips;
 };
 
diff --git a/ctdb/server/ctdb_persistent.c b/ctdb/server/ctdb_persistent.c
index 16e2806546..3f941de606 100644
--- a/ctdb/server/ctdb_persistent.c
+++ b/ctdb/server/ctdb_persistent.c
@@ -91,9 +91,20 @@ int32_t ctdb_control_persistent_store(struct ctdb_context *ctdb,
 	state->c    = c;
 
 	for (i=0;i<ctdb->vnn_map->size;i++) {
-		struct ctdb_node *node = ctdb->nodes[ctdb->vnn_map->map[i]];
+		uint32_t nodeid;
+		struct ctdb_node *node;
 		int ret;
 
+
+		nodeid = ctdb->vnn_map->map[i];
+		if ( !ctdb_validate_pnn(ctdb, nodeid) ) {
+			DEBUG(0, ("Cant update persistent record to non-existent vnn node:%u.\n", nodeid));
+			continue;
+		}
+
+
+		node = ctdb->nodes[nodeid];
+
 		/* only send to active nodes */
 		if (node->flags & NODE_FLAGS_INACTIVE) {
 			continue;
diff --git a/ctdb/server/ctdb_recoverd.c b/ctdb/server/ctdb_recoverd.c
index 9c5fdda954..cd75bb82b8 100644
--- a/ctdb/server/ctdb_recoverd.c
+++ b/ctdb/server/ctdb_recoverd.c
@@ -1786,6 +1786,8 @@ again:
 	   they are the same as for this node
 	 */
 	for (j=0; j<nodemap->num; j++) {
+		uint32_t node_count;
+
 		if (nodemap->nodes[j].flags & NODE_FLAGS_INACTIVE) {
 			continue;
 		}
@@ -1801,20 +1803,34 @@ again:
 			goto again;
 		}
 
-		/* if the nodes disagree on how many nodes there are
-		   then this is a good reason to try recovery
+		/* If the nodes disagree on how many nodes there are
+		   then this is a good reason to try recovery.
+		   This check might be disabled, for example when adding nodes
+		   to an existing cluster.
 		 */
 		if (remote_nodemap->num != nodemap->num) {
 			DEBUG(0, (__location__ " Remote node:%u has different node count. %u vs %u of the local node\n",
 				  nodemap->nodes[j].pnn, remote_nodemap->num, nodemap->num));
-			do_recovery(rec, mem_ctx, pnn, num_active, nodemap, vnnmap, nodemap->nodes[j].pnn);
-			goto again;
+
+			/* If we have disabled checking the nodes file we just
+			   print a message and continue.
+			 */
+			if (ctdb->tunable.check_nodes_file == 0) {
+				DEBUG(0, ("Node file checking is disabled. Ignoring that node-count is inconsistent\n"));
+			} else {
+				do_recovery(rec, mem_ctx, pnn, num_active, nodemap, vnnmap, nodemap->nodes[j].pnn);
+				goto again;
+			}
 		}
 
 		/* if the nodes disagree on which nodes exist and are
 		   active, then that is also a good reason to do recovery
 		 */
-		for (i=0;i<nodemap->num;i++) {
+		node_count = nodemap->num;
+		if (remote_nodemap->num < node_count) {
+			node_count = remote_nodemap->num;
+		}
+		for (i=0;i<node_count;i++) {
 			if (remote_nodemap->nodes[i].pnn != nodemap->nodes[i].pnn) {
 				DEBUG(0, (__location__ " Remote node:%u has different nodemap pnn for %d (%u vs %u).\n", 
 					  nodemap->nodes[j].pnn, i, 
diff --git a/ctdb/server/ctdb_server.c b/ctdb/server/ctdb_server.c
index dddf90753b..497ae1f695 100644
--- a/ctdb/server/ctdb_server.c
+++ b/ctdb/server/ctdb_server.c
@@ -204,7 +204,17 @@ uint32_t ctdb_get_num_active_nodes(struct ctdb_context *ctdb)
 	int i;
 	uint32_t count=0;
 	for (i=0;i<ctdb->vnn_map->size;i++) {
-		struct ctdb_node *node = ctdb->nodes[ctdb->vnn_map->map[i]];
+		uint32_t nodeid;
+		struct ctdb_node *node;
+
+		nodeid = ctdb->vnn_map->map[i];
+		if ( !ctdb_validate_pnn(ctdb, nodeid) ) {
+			DEBUG(0, ("Ignoring non-existent vnn node:%u when counting number of active nodes.\n", nodeid));
+			continue;
+		}
+
+
+		node = ctdb->nodes[nodeid];
 		if (!(node->flags & NODE_FLAGS_INACTIVE)) {
 			count++;
 		}
diff --git a/ctdb/server/ctdb_takeover.c b/ctdb/server/ctdb_takeover.c
index ec3455e4c0..324ab70370 100644
--- a/ctdb/server/ctdb_takeover.c
+++ b/ctdb/server/ctdb_takeover.c
@@ -694,6 +694,12 @@ int ctdb_takeover_run(struct ctdb_context *ctdb, struct ctdb_node_map *nodemap)
 		if (tmp_ip->pnn == -1) {
 			continue;
 		}
+
+		if (!ctdb_validate_pnn(ctdb, tmp_ip->pnn)) {
+			tmp_ip->pnn = -1;
+			continue;
+		}
+
 		if (nodemap->nodes[tmp_ip->pnn].flags & mask) {
 			tmp_ip->pnn = -1;
 		}
diff --git a/ctdb/server/ctdb_tunables.c b/ctdb/server/ctdb_tunables.c
index da1f6e6ba7..f738ae3260 100644
--- a/ctdb/server/ctdb_tunables.c
+++ b/ctdb/server/ctdb_tunables.c
@@ -44,6 +44,7 @@ static const struct {
 	{ "DatabaseHashSize", 10000,  offsetof(struct ctdb_tunable, database_hash_size) },
 	{ "RerecoveryTimeout",   10,  offsetof(struct ctdb_tunable, rerecovery_timeout) },
 	{ "EnableBans",           1,  offsetof(struct ctdb_tunable, enable_bans) },
+	{ "CheckNodesFile",       1,  offsetof(struct ctdb_tunable, check_nodes_file) },
 	{ "DeterministicIPs",     0,  offsetof(struct ctdb_tunable, deterministic_public_ips) },
 };
 
-- 
cgit 


From 3c1f9882a804c7e78704f41b6e2765a1fe938fa0 Mon Sep 17 00:00:00 2001
From: Ronnie Sahlberg <sahlberg@ronnie>
Date: Mon, 12 Nov 2007 10:23:35 +1100
Subject: revert 773

(This used to be ctdb commit 5a1c8f458ddc9b0ff532afda6007e32db10a71c8)
---
 ctdb/include/ctdb_private.h   |  1 -
 ctdb/server/ctdb_persistent.c | 13 +------------
 ctdb/server/ctdb_recoverd.c   | 26 +++++---------------------
 ctdb/server/ctdb_server.c     | 12 +-----------
 ctdb/server/ctdb_takeover.c   |  6 ------
 ctdb/server/ctdb_tunables.c   |  1 -
 6 files changed, 7 insertions(+), 52 deletions(-)

(limited to 'ctdb')

diff --git a/ctdb/include/ctdb_private.h b/ctdb/include/ctdb_private.h
index 9df80b64af..7b98683e9f 100644
--- a/ctdb/include/ctdb_private.h
+++ b/ctdb/include/ctdb_private.h
@@ -88,7 +88,6 @@ struct ctdb_tunable {
 	uint32_t database_hash_size;
 	uint32_t rerecovery_timeout;
 	uint32_t enable_bans;
-	uint32_t check_nodes_file;
 	uint32_t deterministic_public_ips;
 };
 
diff --git a/ctdb/server/ctdb_persistent.c b/ctdb/server/ctdb_persistent.c
index 3f941de606..16e2806546 100644
--- a/ctdb/server/ctdb_persistent.c
+++ b/ctdb/server/ctdb_persistent.c
@@ -91,20 +91,9 @@ int32_t ctdb_control_persistent_store(struct ctdb_context *ctdb,
 	state->c    = c;
 
 	for (i=0;i<ctdb->vnn_map->size;i++) {
-		uint32_t nodeid;
-		struct ctdb_node *node;
+		struct ctdb_node *node = ctdb->nodes[ctdb->vnn_map->map[i]];
 		int ret;
 
-
-		nodeid = ctdb->vnn_map->map[i];
-		if ( !ctdb_validate_pnn(ctdb, nodeid) ) {
-			DEBUG(0, ("Cant update persistent record to non-existent vnn node:%u.\n", nodeid));
-			continue;
-		}
-
-
-		node = ctdb->nodes[nodeid];
-
 		/* only send to active nodes */
 		if (node->flags & NODE_FLAGS_INACTIVE) {
 			continue;
diff --git a/ctdb/server/ctdb_recoverd.c b/ctdb/server/ctdb_recoverd.c
index cd75bb82b8..9c5fdda954 100644
--- a/ctdb/server/ctdb_recoverd.c
+++ b/ctdb/server/ctdb_recoverd.c
@@ -1786,8 +1786,6 @@ again:
 	   they are the same as for this node
 	 */
 	for (j=0; j<nodemap->num; j++) {
-		uint32_t node_count;
-
 		if (nodemap->nodes[j].flags & NODE_FLAGS_INACTIVE) {
 			continue;
 		}
@@ -1803,34 +1801,20 @@ again:
 			goto again;
 		}
 
-		/* If the nodes disagree on how many nodes there are
-		   then this is a good reason to try recovery.
-		   This check might be disabled, for example when adding nodes
-		   to an existing cluster.
+		/* if the nodes disagree on how many nodes there are
+		   then this is a good reason to try recovery
 		 */
 		if (remote_nodemap->num != nodemap->num) {
 			DEBUG(0, (__location__ " Remote node:%u has different node count. %u vs %u of the local node\n",
 				  nodemap->nodes[j].pnn, remote_nodemap->num, nodemap->num));
-
-			/* If we have disabled checking the nodes file we just
-			   print a message and continue.
-			 */
-			if (ctdb->tunable.check_nodes_file == 0) {
-				DEBUG(0, ("Node file checking is disabled. Ignoring that node-count is inconsistent\n"));
-			} else {
-				do_recovery(rec, mem_ctx, pnn, num_active, nodemap, vnnmap, nodemap->nodes[j].pnn);
-				goto again;
-			}
+			do_recovery(rec, mem_ctx, pnn, num_active, nodemap, vnnmap, nodemap->nodes[j].pnn);
+			goto again;
 		}
 
 		/* if the nodes disagree on which nodes exist and are
 		   active, then that is also a good reason to do recovery
 		 */
-		node_count = nodemap->num;
-		if (remote_nodemap->num < node_count) {
-			node_count = remote_nodemap->num;
-		}
-		for (i=0;i<node_count;i++) {
+		for (i=0;i<nodemap->num;i++) {
 			if (remote_nodemap->nodes[i].pnn != nodemap->nodes[i].pnn) {
 				DEBUG(0, (__location__ " Remote node:%u has different nodemap pnn for %d (%u vs %u).\n", 
 					  nodemap->nodes[j].pnn, i, 
diff --git a/ctdb/server/ctdb_server.c b/ctdb/server/ctdb_server.c
index 497ae1f695..dddf90753b 100644
--- a/ctdb/server/ctdb_server.c
+++ b/ctdb/server/ctdb_server.c
@@ -204,17 +204,7 @@ uint32_t ctdb_get_num_active_nodes(struct ctdb_context *ctdb)
 	int i;
 	uint32_t count=0;
 	for (i=0;i<ctdb->vnn_map->size;i++) {
-		uint32_t nodeid;
-		struct ctdb_node *node;
-
-		nodeid = ctdb->vnn_map->map[i];
-		if ( !ctdb_validate_pnn(ctdb, nodeid) ) {
-			DEBUG(0, ("Ignoring non-existent vnn node:%u when counting number of active nodes.\n", nodeid));
-			continue;
-		}
-
-
-		node = ctdb->nodes[nodeid];
+		struct ctdb_node *node = ctdb->nodes[ctdb->vnn_map->map[i]];
 		if (!(node->flags & NODE_FLAGS_INACTIVE)) {
 			count++;
 		}
diff --git a/ctdb/server/ctdb_takeover.c b/ctdb/server/ctdb_takeover.c
index 324ab70370..ec3455e4c0 100644
--- a/ctdb/server/ctdb_takeover.c
+++ b/ctdb/server/ctdb_takeover.c
@@ -694,12 +694,6 @@ int ctdb_takeover_run(struct ctdb_context *ctdb, struct ctdb_node_map *nodemap)
 		if (tmp_ip->pnn == -1) {
 			continue;
 		}
-
-		if (!ctdb_validate_pnn(ctdb, tmp_ip->pnn)) {
-			tmp_ip->pnn = -1;
-			continue;
-		}
-
 		if (nodemap->nodes[tmp_ip->pnn].flags & mask) {
 			tmp_ip->pnn = -1;
 		}
diff --git a/ctdb/server/ctdb_tunables.c b/ctdb/server/ctdb_tunables.c
index f738ae3260..da1f6e6ba7 100644
--- a/ctdb/server/ctdb_tunables.c
+++ b/ctdb/server/ctdb_tunables.c
@@ -44,7 +44,6 @@ static const struct {
 	{ "DatabaseHashSize", 10000,  offsetof(struct ctdb_tunable, database_hash_size) },
 	{ "RerecoveryTimeout",   10,  offsetof(struct ctdb_tunable, rerecovery_timeout) },
 	{ "EnableBans",           1,  offsetof(struct ctdb_tunable, enable_bans) },
-	{ "CheckNodesFile",       1,  offsetof(struct ctdb_tunable, check_nodes_file) },
 	{ "DeterministicIPs",     0,  offsetof(struct ctdb_tunable, deterministic_public_ips) },
 };
 
-- 
cgit 


From dfa6829621adddb1fdcca1d3e067ec7fee5e473c Mon Sep 17 00:00:00 2001
From: Ronnie Sahlberg <sahlberg@ronnie>
Date: Wed, 14 Nov 2007 16:17:52 +1100
Subject: add CTDB_MANAGES_WINBIND to /etc/sysconfig/ctdb to allow ctdb to be
 used in environments where samba is used without winbind

(This used to be ctdb commit 1ae5af14f90fd81a20b14c02c0c5ad355a609134)
---
 ctdb/config/ctdb.sysconfig    |  3 +++
 ctdb/config/events.d/50.samba | 26 ++++++++++++++++++--------
 ctdb/web/samba.html           | 17 ++++++++++++++++-
 3 files changed, 37 insertions(+), 9 deletions(-)

(limited to 'ctdb')

diff --git a/ctdb/config/ctdb.sysconfig b/ctdb/config/ctdb.sysconfig
index d290321b99..bbb6f270b1 100644
--- a/ctdb/config/ctdb.sysconfig
+++ b/ctdb/config/ctdb.sysconfig
@@ -33,6 +33,9 @@
 # default is to not manage Samba
 # CTDB_MANAGES_SAMBA=yes
 
+# should ctdb manage starting/stopping Winbind service?
+# CTDB_MANAGES_WINBIND=yes
+
 # you may wish to raise the file descriptor limit for ctdb
 # use a ulimit command here. ctdb needs one file descriptor per
 # connected client (ie. one per connected client in Samba)
diff --git a/ctdb/config/events.d/50.samba b/ctdb/config/events.d/50.samba
index 5f9fa4fafa..88d8e4e438 100755
--- a/ctdb/config/events.d/50.samba
+++ b/ctdb/config/events.d/50.samba
@@ -32,17 +32,21 @@ case $cmd in
 
 	# make sure samba is not already started
 	service smb stop > /dev/null 2>&1
-	service winbind stop > /dev/null 2>&1
-	killall -0 -q smbd winbindd && {
-	    sleep 1
-	    # make absolutely sure samba is dead
-	    killall -q -9 smbd winbindd
+
+	# restart the winbind service
+	[ "$CTDB_MANAGES_WINBIND" = "yes" ] && {
+		service winbind stop > /dev/null 2>&1
+		killall -0 -q smbd winbindd && {
+		    sleep 1
+		    # make absolutely sure samba is dead
+		    killall -q -9 smbd winbindd
+		}
+		service winbind start
 	}
 
 	# start Samba service. Start it reniced, as under very heavy load 
 	# the number of smbd processes will mean that it leaves few cycles for
 	# anything else
-	service winbind start
 	nice service smb start
 	;;
 	
@@ -62,7 +66,11 @@ case $cmd in
      shutdown)
 	# shutdown Samba when ctdb goes down
 	service smb stop
-	service winbind stop
+
+	# stop the winbind service
+	[ "$CTDB_MANAGES_WINBIND" = "yes" ] && {
+		service winbind stop
+	}
 	;;
 
      monitor)
@@ -89,7 +97,9 @@ case $cmd in
 	ctdb_check_tcp_ports "Samba" $smb_ports
 
 	# check winbind is OK
-	ctdb_check_command "winbind" "wbinfo -p"
+	[ "$CTDB_MANAGES_WINBIND" = "yes" ] && {
+		ctdb_check_command "winbind" "wbinfo -p"
+	}
 	;;
 
 esac
diff --git a/ctdb/web/samba.html b/ctdb/web/samba.html
index 0b5be7f6bc..08fc683ab8 100644
--- a/ctdb/web/samba.html
+++ b/ctdb/web/samba.html
@@ -64,7 +64,7 @@ A suitable file can be found in the dbench distribution at http://samba.org/ftp/
 
 <h3>CTDB_MANAGES_SAMBA</h3>
 This is a parameter in /etc/sysconfig/ctdb<br><br>
-When this parameter is set to "yes" CTDB will start/stop/restart the localo samba daemon as the cluster configuration changes.<br><br>
+When this parameter is set to "yes" CTDB will start/stop/restart the local samba daemon as the cluster configuration changes.<br><br>
 When this parameter is set you should also make sure that samba is NOT started by default by the linux system when it boots:
 <pre>
   chkconfig samba off
@@ -77,5 +77,20 @@ Example:
 
 It is strongly recommended that you set this parameter to "yes" if you intend to use clustered samba.
 
+<h3>CTDB_MANAGES_WINBIND</h3>
+This is a parameter in /etc/sysconfig/ctdb<br><br>
+When this parameter is set to "yes" CTDB will start/stop/restart the local winbind daemon as the cluster configuration changes.<br><br>
+When this parameter is set you should also make sure that winbind is NOT started by default by the linux system when it boots:
+<pre>
+  chkconfig winbind off
+</pre>
+
+Example:
+<pre>
+  CTDB_MANAGES_WINBIND="yes"
+</pre>
+
+It is strongly recommended that you set this parameter to "yes" if you intend to use clustered samba.
+
 <!--#include virtual="footer.html" -->
 
-- 
cgit 


From 9f4b0dab03d832407e6d573c10c2645bd25b335d Mon Sep 17 00:00:00 2001
From: Ronnie Sahlberg <sahlberg@ronnie>
Date: Thu, 15 Nov 2007 06:56:02 +1100
Subject: only check port 21 when monitoring vsftpd

(This used to be ctdb commit 41b0d71aaee186138eddc97d49503841fa26f234)
---
 ctdb/config/events.d/40.vsftpd | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'ctdb')

diff --git a/ctdb/config/events.d/40.vsftpd b/ctdb/config/events.d/40.vsftpd
index d5f8541bb7..bc420fd42d 100755
--- a/ctdb/config/events.d/40.vsftpd
+++ b/ctdb/config/events.d/40.vsftpd
@@ -42,7 +42,7 @@ case $cmd in
 	;;
 
      monitor)
-	ctdb_check_tcp_ports "ftp" 20 21
+	ctdb_check_tcp_ports "ftp" 21
 	;;
 esac
 
-- 
cgit 


From b09d3de7596054689cb18fdbd642ced63059e075 Mon Sep 17 00:00:00 2001
From: Ronnie Sahlberg <sahlberg@ronnie>
Date: Fri, 16 Nov 2007 13:37:27 +1100
Subject: from Christian A when monitoring that all nfs shares are available,
 allow both ' ' and '\t' characters to separate the exported directory from
 the options in /etc/exports

(This used to be ctdb commit ac6cfe9de0acdcf9461068684fa890504454aae4)
---
 ctdb/config/events.d/60.nfs | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

(limited to 'ctdb')

diff --git a/ctdb/config/events.d/60.nfs b/ctdb/config/events.d/60.nfs
index 6c04f1e303..ed9d6da63b 100755
--- a/ctdb/config/events.d/60.nfs
+++ b/ctdb/config/events.d/60.nfs
@@ -69,7 +69,7 @@ case $cmd in
 	ctdb_check_rpc "NFS" 100003 3
 
 	# and that its directories are available
-	nfs_dirs=`grep -v '^#' < /etc/exports | cut -d' ' -f1`
+	nfs_dirs=$(grep -v '^#' < /etc/exports | awk {'print $1;'})
 	ctdb_check_directories "nfs" $nfs_dirs
 
 	# check that lockd responds to rpc requests
-- 
cgit