glusterd: Allocate fresh port on brick (re)start

There is no point of using the same port through the entire volume life cycle
for a particular bricks process since there is no guarantee that the same port
would be free and no other application wouldn't consume it in between the
glusterd/volume restart.

We hit a race where on glusterd restart the daemon services start followed by
brick processes and the time brick process tries to bind with the port which was
allocated by glusterd before a restart is been already consumed by some other
client like NFS/SHD/...

Note : This is a short term solution as here we reduce the race window but don't
eliminate it completely. As a long term solution the port allocation has to be
done by glusterfsd and the same should be communicated back to glusterd for book
keeping

Change-Id: Ibbd1e7ca87e51a7cd9cf216b1fe58ef7783aef24
BUG: 1322805
Signed-off-by: Atin Mukherjee <amukherj@redhat.com>
Reviewed-on: http://review.gluster.org/13865
Smoke: Gluster Build System <jenkins@build.gluster.com>
NetBSD-regression: NetBSD Build System <jenkins@build.gluster.org>
CentOS-regression: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Jeff Darcy <jdarcy@redhat.com>

5 files changed, 46 insertions, 70 deletions

diff --git a/tests/bugs/glusterd/1313628-import-brick-ports-always.t b/tests/bugs/glusterd/1313628-import-brick-ports-always.t
deleted file mode 100755
index d04c4293466..00000000000
--- a/tests/bugs/glusterd/1313628-import-brick-ports-always.t
+++ /dev/null
@@ -1,47 +0,0 @@
-#!/bin/bash
-. $(dirname $0)/../../include.rc
-. $(dirname $0)/../../cluster.rc
-
-## Check that brick ports are always copied on import
-## --------------------------------------------------
-## This test checks that the brick ports are copied on import by checking that
-## they don't change when the following happens,
-##  - Stop a volume
-##  - Stop glusterd
-##  - Start the stopped volume
-##  - Start the stopped glusterd
-
-function get_brick_port() {
-  local VOL=$1
-  local BRICK=$2
-  $CLI2 volume status $VOL $BRICK --xml | sed -ne 's/.*<port>\([0-9]*\)<\/port>/\1/p'
-}
-
-
-cleanup
-
-TEST launch_cluster 2
-TEST $CLI1 peer probe $H2
-EXPECT_WITHIN $PROBE_TIMEOUT 1 peer_count
-
-# Create and start volume so that brick port assignment happens
-TEST $CLI1 volume create $V0 $H1:$B1/$V0 $H2:$B2/$V0
-TEST $CLI1 volume start $V0
-
-# Save port for 2nd brick
-BPORT_ORIG=$(get_brick_port $V0 $H2:$B2/$V0)
-
-# Stop volume, stop 2nd glusterd, start volume, start 2nd glusterd
-TEST $CLI1 volume stop $V0
-TEST kill_glusterd 2
-
-TEST $CLI1 volume start $V0
-TEST start_glusterd 2
-EXPECT_WITHIN $PROBE_TIMEOUT 1 peer_count
-
-# Get new port and compare with old one
-EXPECT_WITHIN $PROCESS_UP_TIMEOUT $BPORT_ORIG get_brick_port $V0 $H2:$B2/$V0
-
-$CLI1 volume stop $V0
-
-cleanup
diff --git a/tests/features/ssl-ciphers.t b/tests/features/ssl-ciphers.t
index 9ee7fc6c16f..f5909f320ac 100644
--- a/tests/features/ssl-ciphers.t
+++ b/tests/features/ssl-ciphers.t
@@ -137,6 +137,7 @@ EXPECT "`pwd`/`dirname $0`/dh1024.pem" volume_option $V0 ssl.dh-param
 TEST $CLI volume stop $V0
 TEST $CLI volume start $V0
 EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" online_brick_count
+BRICK_PORT=`brick_port $V0`
 EXPECT "Y" openssl_connect -cipher EDH -connect $H0:$BRICK_PORT
 
 # Test the cipher-list option
@@ -145,6 +146,7 @@ EXPECT AES256-SHA volume_option $V0 ssl.cipher-list
 TEST $CLI volume stop $V0
 TEST $CLI volume start $V0
 EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" online_brick_count
+BRICK_PORT=`brick_port $V0`
 EXPECT "Y" openssl_connect -cipher AES256-SHA -connect $H0:$BRICK_PORT
 EXPECT "N" openssl_connect -cipher AES128-SHA -connect $H0:$BRICK_PORT
 
@@ -154,6 +156,7 @@ EXPECT EECDH:EDH:!TLSv1 volume_option $V0 ssl.cipher-list
 TEST $CLI volume stop $V0
 TEST $CLI volume start $V0
 EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" online_brick_count
+BRICK_PORT=`brick_port $V0`
 EXPECT "N" openssl_connect -cipher AES256-SHA -connect $H0:$BRICK_PORT
 EXPECT "Y" openssl_connect -cipher EECDH -connect $H0:$BRICK_PORT
 
@@ -162,6 +165,7 @@ EXPECT invalid volume_option $V0 ssl.ec-curve
 TEST $CLI volume stop $V0
 TEST $CLI volume start $V0
 EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" online_brick_count
+BRICK_PORT=`brick_port $V0`
 EXPECT "N" openssl_connect -cipher EECDH -connect $H0:$BRICK_PORT
 
 TEST $CLI volume set $V0 ssl.ec-curve secp521r1
@@ -169,6 +173,7 @@ EXPECT secp521r1 volume_option $V0 ssl.ec-curve
 TEST $CLI volume stop $V0
 TEST $CLI volume start $V0
 EXPECT_WITHIN $CHILD_UP_TIMEOUT "1" online_brick_count
+BRICK_PORT=`brick_port $V0`
 EXPECT "Y" openssl_connect -cipher EECDH -connect $H0:$BRICK_PORT
 
 # test revocation
diff --git a/xlators/mgmt/glusterd/src/glusterd-messages.h b/xlators/mgmt/glusterd/src/glusterd-messages.h
index ffcfa7773e9..04d630001df 100644
--- a/xlators/mgmt/glusterd/src/glusterd-messages.h
+++ b/xlators/mgmt/glusterd/src/glusterd-messages.h
@@ -41,7 +41,7 @@
 
 #define GLUSTERD_COMP_BASE      GLFS_MSGID_GLUSTERD
 
-#define GLFS_NUM_MESSAGES       573
+#define GLFS_NUM_MESSAGES       574
 
 #define GLFS_MSGID_END          (GLUSTERD_COMP_BASE + GLFS_NUM_MESSAGES + 1)
 /* Messaged with message IDs */
@@ -4633,6 +4633,14 @@
  */
 #define GD_MSG_SNAP_RESUME_TIER_FAIL                (GLUSTERD_COMP_BASE + 573)
 
+/*!
+ * @messageid
+ * @diagnosis
+ * @recommendedaction
+ *
+ */
+#define GD_MSG_PMAP_REMOVE_FAIL                (GLUSTERD_COMP_BASE + 574)
+
 /*------------*/
 #define glfs_msg_end_x GLFS_MSGID_END, "Invalid: End of messages"
 #endif /* !_GLUSTERD_MESSAGES_H_ */
diff --git a/xlators/mgmt/glusterd/src/glusterd-store.c b/xlators/mgmt/glusterd/src/glusterd-store.c
index b1f943074f3..97f5147dd39 100644
--- a/xlators/mgmt/glusterd/src/glusterd-store.c
+++ b/xlators/mgmt/glusterd/src/glusterd-store.c
@@ -2313,13 +2313,6 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo)
                                         /* This is required to adhere to the
                                            IANA standards */
                                         brickinfo->port = 0;
-                                } else {
-                                        /* This is required to have proper ports
-                                           assigned to bricks after restart */
-                                        pmap = pmap_registry_get (THIS);
-                                        if (pmap->last_alloc <= brickinfo->port)
-                                                pmap->last_alloc =
-                                                        brickinfo->port + 1;
                                 }
                         } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_RDMA_PORT,
                                     strlen (GLUSTERD_STORE_KEY_BRICK_RDMA_PORT))) {
@@ -2329,16 +2322,7 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo)
                                         /* This is required to adhere to the
                                            IANA standards */
                                         brickinfo->rdma_port = 0;
-                                } else {
-                                        /* This is required to have proper ports
-                                           assigned to bricks after restart */
-                                        pmap = pmap_registry_get (THIS);
-                                        if (pmap->last_alloc <=
-                                            brickinfo->rdma_port)
-                                                pmap->last_alloc =
-                                                        brickinfo->rdma_port +1;
                                 }
-
                         } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED,
                                              strlen (GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED))) {
                                 gf_string2int (value, &brickinfo->decommissioned);
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index a0cc5d409c8..55597cd39cf 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -1725,9 +1725,22 @@ glusterd_volume_start_glusterfs (glusterd_volinfo_t  *volinfo,
         if (gf_is_service_running (pidfile, NULL))
                 goto connect;
 
-        port = brickinfo->port;
-        if (!port)
-                port = pmap_registry_alloc (THIS);
+        /* Do a pmap registry remove on the older connected port */
+        if (brickinfo->port) {
+                ret = pmap_registry_remove (this, brickinfo->port,
+                                            brickinfo->path,
+                                            GF_PMAP_PORT_BRICKSERVER,
+                                            NULL);
+                if (ret) {
+                        gf_msg (this->name, GF_LOG_ERROR, 0,
+                                GD_MSG_PMAP_REMOVE_FAIL, "Failed to remove pmap"
+                                "registry for older signin port %d",
+                                brickinfo->port);
+                        goto out;
+                }
+        }
+
+        port = pmap_registry_alloc (THIS);
 
         /* Build the exp_path, before starting the glusterfsd even in
            valgrind mode. Otherwise all the glusterfsd processes start
@@ -1790,9 +1803,22 @@ glusterd_volume_start_glusterfs (glusterd_volinfo_t  *volinfo,
         if (volinfo->transport_type != GF_TRANSPORT_BOTH_TCP_RDMA) {
                 runner_argprintf (&runner, "%d", port);
         } else {
-                rdma_port = brickinfo->rdma_port;
-                if (!rdma_port)
-                        rdma_port = pmap_registry_alloc (THIS);
+                /* Do a pmap registry remove on the older connected port */
+                if (brickinfo->rdma_port) {
+                        ret = pmap_registry_remove (this, brickinfo->rdma_port,
+                                                    brickinfo->path,
+                                                    GF_PMAP_PORT_BRICKSERVER,
+                                                    NULL);
+                        if (ret) {
+                                gf_msg (this->name, GF_LOG_ERROR, 0,
+                                        GD_MSG_PMAP_REMOVE_FAIL, "Failed to "
+                                        "remove pmap registry for older signin "
+                                        "port %d", brickinfo->rdma_port);
+                                goto out;
+                        }
+                }
+                rdma_port = pmap_registry_alloc (THIS);
+
                 runner_argprintf (&runner, "%d,%d", port, rdma_port);
                 runner_add_arg (&runner, "--xlator-option");
                 runner_argprintf (&runner, "%s-server.transport.rdma.listen-port=%d",