glusterd: allow multiple instances of glusterd on one machine

This is needed to support automated testing of cluster-communication
features such as probing and quorum.  In order to use this, you need to
do the following preparatory steps.

* Copy /var/lib/glusterd to another directory for each virtual host

* Ensure that each virtual host has a different UUID in its glusterd.info

Now you can start each copy of glusterd with the following xlator-options.

* management.transport.socket.bind-address=$ip_address

* management.working-directory=$unique_working_directory

You can use 127.x.y.z addresses for binding without needing to assign
them to interfaces explicitly.  Note that you must use addresses, not
names, because of some stuff in the socket code that's not worth fixing
just for this usage, but after that you can use names in /etc/hosts
instead.

At this point you can issue CLI commands to a specific glusterd using
the --remote-host option.  So far probe, volume create/start/stop,
mount, and basic I/O all seem to work as expected with multiple
instances.

Change-Id: I1beabb44cff8763d2774bc208b2ffcda27c1a550
BUG: 913555
Signed-off-by: Jeff Darcy <jdarcy@redhat.com>
Reviewed-on: http://review.gluster.org/4556
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>

4 files changed, 142 insertions, 1 deletions

diff --git a/tests/bugs/bug-913555.t b/tests/bugs/bug-913555.t
new file mode 100755
index 00000000000..0e08bd377ae
--- /dev/null
+++ b/tests/bugs/bug-913555.t
@@ -0,0 +1,66 @@
+#!/bin/bash
+
+# Test that a volume becomes unwritable when the cluster loses quorum.
+
+. $(dirname $0)/../include.rc
+. $(dirname $0)/../volume.rc
+
+function vglusterd {
+	wd=$1/wd-$2
+	cp -r /var/lib/glusterd $wd
+	rm -rf $wd/peers/* $wd/vols/*
+	echo -n "UUID=$(uuidgen)\noperating-version=1\n" > $wd/glusterd.info
+	opt1="management.transport.socket.bind-address=127.0.0.$2"
+	opt2="management.working-directory=$wd"
+	glusterd --xlator-option $opt1 --xlator-option $opt2
+}
+
+function check_fs {
+	df $1 &> /dev/null
+	echo $?
+}
+
+function check_peers {
+	$VCLI peer status | grep 'Peer in Cluster (Connected)' | wc -l
+}
+
+cleanup;
+
+topwd=$(mktemp -d)
+trap "rm -rf $topwd" EXIT
+
+vglusterd $topwd 100
+VCLI="$CLI --remote-host=127.0.0.100"
+vglusterd $topwd 101
+TEST $VCLI peer probe 127.0.0.101
+vglusterd $topwd 102
+TEST $VCLI peer probe 127.0.0.102
+
+EXPECT_WITHIN 20 2 check_peers
+
+create_cmd="$VCLI volume create $V0"
+for i in $(seq 100 102); do
+	mkdir -p $B0/$V0$i
+	create_cmd="$create_cmd 127.0.0.$i:$B0/$V0$i"
+done
+
+TEST $create_cmd
+TEST $VCLI volume set $V0 cluster.server-quorum-type server
+TEST $VCLI volume start $V0
+TEST glusterfs --volfile-server=127.0.0.100 --volfile-id=$V0 $M0
+
+# Kill one pseudo-node, make sure the others survive and volume stays up.
+kill -9 $(ps -ef | grep gluster | grep 127.0.0.102 | awk '{print $2}')
+EXPECT_WITHIN 20 1 check_peers
+fs_status=$(check_fs $M0)
+nnodes=$(pidof glusterfsd | wc -w)
+TEST [ "$fs_status" = 0 -a "$nnodes" = 2 ]
+
+# Kill another pseudo-node, make sure the last one dies and volume goes down.
+kill -9 $(ps -ef | grep gluster | grep 127.0.0.101 | awk '{print $2}')
+EXPECT_WITHIN 20 0 check_peers
+fs_status=$(check_fs $M0)
+nnodes=$(pidof glusterfsd | wc -w)
+TEST [ "$fs_status" = 1 -a "$nnodes" = 0 ]
+
+cleanup
diff --git a/xlators/mgmt/glusterd/src/glusterd-handler.c b/xlators/mgmt/glusterd/src/glusterd-handler.c
index 7d39c1e0785..04e5833b732 100644
--- a/xlators/mgmt/glusterd/src/glusterd-handler.c
+++ b/xlators/mgmt/glusterd/src/glusterd-handler.c
@@ -707,6 +707,7 @@ glusterd_handle_cli_probe (rpcsvc_request_t *req)
         glusterd_peerinfo_t             *peerinfo = NULL;
         gf_boolean_t                    run_fsm = _gf_true;
         xlator_t                        *this = NULL;
+        char                            *bind_name = NULL;
 
         GF_ASSERT (req);
         this = THIS;
@@ -736,7 +737,16 @@ glusterd_handle_cli_probe (rpcsvc_request_t *req)
         gf_log ("glusterd", GF_LOG_INFO, "Received CLI probe req %s %d",
                 cli_req.hostname, cli_req.port);
 
-        if (glusterd_is_local_addr(cli_req.hostname)) {
+        if (dict_get_str(this->options,"transport.socket.bind-address",
+                         &bind_name) == 0) {
+                gf_log ("glusterd", GF_LOG_DEBUG,
+                        "only checking probe address vs. bind address");
+                ret = glusterd_is_same_address(bind_name,cli_req.hostname);
+        }
+        else {
+                ret = glusterd_is_local_addr(cli_req.hostname);
+        }
+        if (ret) {
                 glusterd_xfer_cli_probe_resp (req, 0, GF_PROBE_LOCALHOST, NULL,
                                               cli_req.hostname, cli_req.port);
                 ret = 0;
@@ -2434,6 +2444,7 @@ glusterd_friend_rpc_create (xlator_t *this, glusterd_peerinfo_t *peerinfo,
         dict_t                 *options = NULL;
         int                    ret = -1;
         glusterd_peerctx_t     *peerctx = NULL;
+        data_t                 *data = NULL;
 
         peerctx = GF_CALLOC (1, sizeof (*peerctx), gf_gld_mt_peerctx_t);
         if (!peerctx)
@@ -2450,6 +2461,19 @@ glusterd_friend_rpc_create (xlator_t *this, glusterd_peerinfo_t *peerinfo,
         if (ret)
                 goto out;
 
+        /*
+         * For simulated multi-node testing, we need to make sure that we
+         * create our RPC endpoint with the same address that the peer would
+         * use to reach us.
+         */
+        if (this->options) {
+                data = dict_get(this->options,"transport.socket.bind-address");
+                if (data) {
+                        ret = dict_set(options,
+                                       "transport.socket.source-addr",data);
+                }
+        }
+
         ret = glusterd_rpc_create (&peerinfo->rpc, options,
                                    glusterd_peer_rpc_notify, peerctx);
         if (ret) {
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index 7d64a2a0810..54c095eec48 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -7396,3 +7396,51 @@ glusterd_copy_uuid_to_dict (uuid_t uuid, dict_t *dict, char *key)
 
         return 0;
 }
+
+gf_boolean_t
+glusterd_is_same_address (char *name1, char *name2)
+{
+        struct addrinfo         *addr1 = NULL;
+        struct addrinfo         *addr2 = NULL;
+        struct addrinfo         *p = NULL;       
+        struct addrinfo         *q = NULL;       
+        gf_boolean_t            ret = _gf_false;
+        int                     gai_err = 0;
+
+        gai_err = getaddrinfo(name1,NULL,NULL,&addr1);
+        if (gai_err != 0) {
+                gf_log (name1, GF_LOG_WARNING,
+                        "error in getaddrinfo: %s\n", gai_strerror(gai_err));
+                goto out;
+        }
+
+        gai_err = getaddrinfo(name2,NULL,NULL,&addr2);
+        if (gai_err != 0) {
+                gf_log (name2, GF_LOG_WARNING,
+                        "error in getaddrinfo: %s\n", gai_strerror(gai_err));
+                goto out;
+        }
+
+        for (p = addr1; p; p = p->ai_next) {
+                for (q = addr2; q; q = q->ai_next) {
+                        if (p->ai_addrlen != q->ai_addrlen) {
+                                continue;
+                        }
+                        if (memcmp(p->ai_addr,q->ai_addr,p->ai_addrlen)) {
+                                continue;
+                        }
+                        ret = _gf_true;
+                        goto out;
+                }
+        }
+
+out:
+        if (addr1) {
+                freeaddrinfo(addr1);
+        }
+        if (addr2) {
+                freeaddrinfo(addr2);
+        }
+        return ret;
+
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.h b/xlators/mgmt/glusterd/src/glusterd-utils.h
index 1f84634582b..b6a8675b223 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.h
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.h
@@ -509,4 +509,7 @@ glusterd_generate_and_set_task_id (dict_t *dict, char *key);
 
 int
 glusterd_copy_uuid_to_dict (uuid_t uuid, dict_t *dict, char *key);
+
+gf_boolean_t
+glusterd_is_same_address (char *name1, char *name2);
 #endif