summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rwxr-xr-xtests/bugs/bug-1173414-mgmt-v3-remote-lock-failure.t34
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-handler.c4
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mem-types.h3
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-mgmt.c168
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-op-sm.c8
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-sm.h5
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-syncop.c170
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-syncop.h21
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-utils.c49
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-utils.h13
10 files changed, 309 insertions, 166 deletions
diff --git a/tests/bugs/bug-1173414-mgmt-v3-remote-lock-failure.t b/tests/bugs/bug-1173414-mgmt-v3-remote-lock-failure.t
new file mode 100755
index 00000000000..adc3fe30dd4
--- /dev/null
+++ b/tests/bugs/bug-1173414-mgmt-v3-remote-lock-failure.t
@@ -0,0 +1,34 @@
+#!/bin/bash
+
+. $(dirname $0)/../include.rc
+. $(dirname $0)/../cluster.rc
+
+function check_peers {
+ $CLI_1 peer status | grep 'Peer in Cluster (Connected)' | wc -l
+}
+
+cleanup;
+
+TEST launch_cluster 2;
+TEST $CLI_1 peer probe $H2;
+
+EXPECT_WITHIN $PROBE_TIMEOUT 1 check_peers
+
+TEST $CLI_1 volume create $V0 $H1:$B1/$V0
+TEST $CLI_1 volume create $V1 $H1:$B1/$V1
+TEST $CLI_1 volume start $V0
+TEST $CLI_1 volume start $V1
+
+for i in {1..20}
+do
+ $CLI_1 volume set $V0 diagnostics.client-log-level DEBUG &
+ $CLI_1 volume set $V1 barrier on
+ $CLI_2 volume set $V0 diagnostics.client-log-level DEBUG &
+ $CLI_2 volume set $V1 barrier on
+done
+
+EXPECT_WITHIN $PROBE_TIMEOUT 1 check_peers
+TEST $CLI_1 volume status
+TEST $CLI_2 volume status
+
+cleanup;
diff --git a/xlators/mgmt/glusterd/src/glusterd-handler.c b/xlators/mgmt/glusterd/src/glusterd-handler.c
index cb10d4dcb3f..5541852f044 100644
--- a/xlators/mgmt/glusterd/src/glusterd-handler.c
+++ b/xlators/mgmt/glusterd/src/glusterd-handler.c
@@ -1001,7 +1001,7 @@ __glusterd_handle_cli_probe (rpcsvc_request_t *req)
}
if (glusterd_is_any_volume_in_server_quorum (this) &&
- !does_gd_meet_server_quorum (this, _gf_false)) {
+ !does_gd_meet_server_quorum (this, NULL, _gf_false)) {
glusterd_xfer_cli_probe_resp (req, -1, GF_PROBE_QUORUM_NOT_MET,
NULL, hostname, port, dict);
gf_msg (this->name, GF_LOG_CRITICAL, 0,
@@ -1167,7 +1167,7 @@ __glusterd_handle_cli_deprobe (rpcsvc_request_t *req)
if (!(flags & GF_CLI_FLAG_OP_FORCE)) {
if (glusterd_is_any_volume_in_server_quorum (this) &&
- !does_gd_meet_server_quorum (this, _gf_false)) {
+ !does_gd_meet_server_quorum (this, NULL, _gf_false)) {
gf_msg (this->name, GF_LOG_CRITICAL, 0,
GD_MSG_SERVER_QUORUM_NOT_MET,
"Server quorum not met. Rejecting operation.");
diff --git a/xlators/mgmt/glusterd/src/glusterd-mem-types.h b/xlators/mgmt/glusterd/src/glusterd-mem-types.h
index 95908cf62fc..ed171b69b66 100644
--- a/xlators/mgmt/glusterd/src/glusterd-mem-types.h
+++ b/xlators/mgmt/glusterd/src/glusterd-mem-types.h
@@ -70,7 +70,8 @@ typedef enum gf_gld_mem_types_ {
gf_gld_mt_snap_t = gf_common_mt_end + 54,
gf_gld_mt_missed_snapinfo_t = gf_common_mt_end + 55,
gf_gld_mt_snap_create_args_t = gf_common_mt_end + 56,
- gf_gld_mt_end = gf_common_mt_end + 57,
+ gf_gld_mt_local_peers_t = gf_common_mt_end + 57,
+ gf_gld_mt_end = gf_common_mt_end + 58,
} gf_gld_mem_types_t;
#endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-mgmt.c b/xlators/mgmt/glusterd/src/glusterd-mgmt.c
index 52b4184ee7a..1c99f92e114 100644
--- a/xlators/mgmt/glusterd/src/glusterd-mgmt.c
+++ b/xlators/mgmt/glusterd/src/glusterd-mgmt.c
@@ -383,28 +383,25 @@ out:
}
int
-glusterd_mgmt_v3_initiate_lockdown (glusterd_conf_t *conf, glusterd_op_t op,
- dict_t *dict, char **op_errstr, int npeers,
- gf_boolean_t *is_acquired)
+glusterd_mgmt_v3_initiate_lockdown (glusterd_op_t op, dict_t *dict,
+ char **op_errstr, int npeers,
+ gf_boolean_t *is_acquired,
+ struct list_head *peers)
{
char *volname = NULL;
glusterd_peerinfo_t *peerinfo = NULL;
int32_t ret = -1;
int32_t peer_cnt = 0;
struct syncargs args = {0};
- struct list_head *peers = NULL;
uuid_t peer_uuid = {0};
xlator_t *this = NULL;
this = THIS;
GF_ASSERT (this);
- GF_ASSERT (conf);
GF_ASSERT (dict);
GF_ASSERT (op_errstr);
GF_ASSERT (is_acquired);
- peers = &conf->xaction_peers;
-
/* Trying to acquire multiple mgmt_v3 locks on local node */
ret = glusterd_multiple_mgmt_v3_lock (dict, MY_UUID);
if (ret) {
@@ -424,7 +421,7 @@ glusterd_mgmt_v3_initiate_lockdown (glusterd_conf_t *conf, glusterd_op_t op,
gd_syncargs_init (&args, NULL);
synctask_barrier_init((&args));
peer_cnt = 0;
- list_for_each_entry (peerinfo, peers, op_peers_list) {
+ list_for_each_local_xaction_peers (peerinfo, peers) {
gd_mgmt_v3_lock (op, dict, peerinfo, &args,
MY_UUID, peer_uuid);
peer_cnt++;
@@ -637,26 +634,23 @@ out:
}
int
-glusterd_mgmt_v3_pre_validate (glusterd_conf_t *conf, glusterd_op_t op,
- dict_t *req_dict, char **op_errstr, int npeers)
+glusterd_mgmt_v3_pre_validate (glusterd_op_t op, dict_t *req_dict,
+ char **op_errstr, int npeers,
+ struct list_head *peers)
{
int32_t ret = -1;
int32_t peer_cnt = 0;
dict_t *rsp_dict = NULL;
glusterd_peerinfo_t *peerinfo = NULL;
struct syncargs args = {0};
- struct list_head *peers = NULL;
uuid_t peer_uuid = {0};
xlator_t *this = NULL;
this = THIS;
GF_ASSERT (this);
- GF_ASSERT (conf);
GF_ASSERT (req_dict);
GF_ASSERT (op_errstr);
- peers = &conf->xaction_peers;
-
rsp_dict = dict_new ();
if (!rsp_dict) {
gf_log (this->name, GF_LOG_ERROR,
@@ -708,7 +702,7 @@ glusterd_mgmt_v3_pre_validate (glusterd_conf_t *conf, glusterd_op_t op,
gd_syncargs_init (&args, req_dict);
synctask_barrier_init((&args));
peer_cnt = 0;
- list_for_each_entry (peerinfo, peers, op_peers_list) {
+ list_for_each_local_xaction_peers (peerinfo, peers) {
gd_mgmt_v3_pre_validate_req (op, req_dict, peerinfo, &args,
MY_UUID, peer_uuid);
peer_cnt++;
@@ -875,26 +869,23 @@ out:
}
int
-glusterd_mgmt_v3_brick_op (glusterd_conf_t *conf, glusterd_op_t op,
- dict_t *req_dict, char **op_errstr, int npeers)
+glusterd_mgmt_v3_brick_op (glusterd_op_t op, dict_t *req_dict,
+ char **op_errstr, int npeers,
+ struct list_head *peers)
{
int32_t ret = -1;
int32_t peer_cnt = 0;
dict_t *rsp_dict = NULL;
glusterd_peerinfo_t *peerinfo = NULL;
struct syncargs args = {0};
- struct list_head *peers = NULL;
uuid_t peer_uuid = {0};
xlator_t *this = NULL;
this = THIS;
GF_ASSERT (this);
- GF_ASSERT (conf);
GF_ASSERT (req_dict);
GF_ASSERT (op_errstr);
- peers = &conf->xaction_peers;
-
rsp_dict = dict_new ();
if (!rsp_dict) {
gf_log (this->name, GF_LOG_ERROR,
@@ -937,7 +928,7 @@ glusterd_mgmt_v3_brick_op (glusterd_conf_t *conf, glusterd_op_t op,
gd_syncargs_init (&args, NULL);
synctask_barrier_init((&args));
peer_cnt = 0;
- list_for_each_entry (peerinfo, peers, op_peers_list) {
+ list_for_each_local_xaction_peers (peerinfo, peers) {
gd_mgmt_v3_brick_op_req (op, req_dict, peerinfo, &args,
MY_UUID, peer_uuid);
peer_cnt++;
@@ -1099,28 +1090,24 @@ out:
}
int
-glusterd_mgmt_v3_commit (glusterd_conf_t *conf, glusterd_op_t op,
- dict_t *op_ctx, dict_t *req_dict,
- char **op_errstr, int npeers)
+glusterd_mgmt_v3_commit (glusterd_op_t op, dict_t *op_ctx,
+ dict_t *req_dict, char **op_errstr,
+ int npeers, struct list_head *peers)
{
int32_t ret = -1;
int32_t peer_cnt = 0;
dict_t *rsp_dict = NULL;
glusterd_peerinfo_t *peerinfo = NULL;
struct syncargs args = {0};
- struct list_head *peers = NULL;
uuid_t peer_uuid = {0};
xlator_t *this = NULL;
this = THIS;
GF_ASSERT (this);
- GF_ASSERT (conf);
GF_ASSERT (op_ctx);
GF_ASSERT (req_dict);
GF_ASSERT (op_errstr);
- peers = &conf->xaction_peers;
-
rsp_dict = dict_new ();
if (!rsp_dict) {
gf_log (this->name, GF_LOG_ERROR,
@@ -1172,7 +1159,7 @@ glusterd_mgmt_v3_commit (glusterd_conf_t *conf, glusterd_op_t op,
gd_syncargs_init (&args, op_ctx);
synctask_barrier_init((&args));
peer_cnt = 0;
- list_for_each_entry (peerinfo, peers, op_peers_list) {
+ list_for_each_local_xaction_peers (peerinfo, peers) {
gd_mgmt_v3_commit_req (op, req_dict, peerinfo, &args,
MY_UUID, peer_uuid);
peer_cnt++;
@@ -1302,29 +1289,24 @@ out:
}
int
-glusterd_mgmt_v3_post_validate (glusterd_conf_t *conf, glusterd_op_t op,
- int32_t op_ret, dict_t *dict, dict_t *req_dict,
- char **op_errstr, int npeers)
+glusterd_mgmt_v3_post_validate (glusterd_op_t op, int32_t op_ret, dict_t *dict,
+ dict_t *req_dict, char **op_errstr, int npeers,
+ struct list_head *peers)
{
int32_t ret = -1;
int32_t peer_cnt = 0;
dict_t *rsp_dict = NULL;
glusterd_peerinfo_t *peerinfo = NULL;
struct syncargs args = {0};
- struct list_head *peers = NULL;
uuid_t peer_uuid = {0};
xlator_t *this = NULL;
this = THIS;
GF_ASSERT (this);
- GF_ASSERT (conf);
GF_ASSERT (dict);
GF_VALIDATE_OR_GOTO (this->name, req_dict, out);
GF_ASSERT (op_errstr);
- peers = &conf->xaction_peers;
- GF_ASSERT (peers);
-
rsp_dict = dict_new ();
if (!rsp_dict) {
gf_log (this->name, GF_LOG_ERROR,
@@ -1370,7 +1352,7 @@ glusterd_mgmt_v3_post_validate (glusterd_conf_t *conf, glusterd_op_t op,
gd_syncargs_init (&args, req_dict);
synctask_barrier_init((&args));
peer_cnt = 0;
- list_for_each_entry (peerinfo, peers, op_peers_list) {
+ list_for_each_local_xaction_peers (peerinfo, peers) {
gd_mgmt_v3_post_validate_req (op, op_ret, req_dict, peerinfo,
&args, MY_UUID, peer_uuid);
peer_cnt++;
@@ -1496,10 +1478,11 @@ out:
}
int
-glusterd_mgmt_v3_release_peer_locks (glusterd_conf_t *conf, glusterd_op_t op,
+glusterd_mgmt_v3_release_peer_locks (glusterd_op_t op,
dict_t *dict, int32_t op_ret,
char **op_errstr, int npeers,
- gf_boolean_t is_acquired)
+ gf_boolean_t is_acquired,
+ struct list_head *peers)
{
int32_t ret = -1;
int32_t peer_cnt = 0;
@@ -1507,16 +1490,12 @@ glusterd_mgmt_v3_release_peer_locks (glusterd_conf_t *conf, glusterd_op_t op,
xlator_t *this = NULL;
glusterd_peerinfo_t *peerinfo = NULL;
struct syncargs args = {0};
- struct list_head *peers = NULL;
this = THIS;
GF_ASSERT (this);
- GF_ASSERT (conf);
GF_ASSERT (dict);
GF_ASSERT (op_errstr);
- peers = &conf->xaction_peers;
-
/* If the lock has not been held during this
* transaction, do not send unlock requests */
if (!is_acquired)
@@ -1531,7 +1510,7 @@ glusterd_mgmt_v3_release_peer_locks (glusterd_conf_t *conf, glusterd_op_t op,
gd_syncargs_init (&args, NULL);
synctask_barrier_init((&args));
peer_cnt = 0;
- list_for_each_entry (peerinfo, peers, op_peers_list) {
+ list_for_each_local_xaction_peers (peerinfo, peers) {
gd_mgmt_v3_unlock (op, dict, peerinfo, &args,
MY_UUID, peer_uuid);
peer_cnt++;
@@ -1569,6 +1548,7 @@ glusterd_mgmt_v3_initiate_all_phases (rpcsvc_request_t *req, glusterd_op_t op,
xlator_t *this = NULL;
gf_boolean_t is_acquired = _gf_false;
uuid_t *originator_uuid = NULL;
+ struct list_head xaction_peers = {0,};
this = THIS;
GF_ASSERT (this);
@@ -1577,6 +1557,15 @@ glusterd_mgmt_v3_initiate_all_phases (rpcsvc_request_t *req, glusterd_op_t op,
conf = this->private;
GF_ASSERT (conf);
+ INIT_LIST_HEAD (&xaction_peers);
+ npeers = gd_build_local_xaction_peers_list (&conf->peers,
+ &xaction_peers, op);
+ if (npeers == -1) {
+ gf_log (this->name, GF_LOG_ERROR, "building local peers list "
+ "failed");
+ goto rsp;
+ }
+
/* Save the MY_UUID as the originator_uuid. This originator_uuid
* will be used by is_origin_glusterd() to determine if a node
* is the originator node for a command. */
@@ -1613,13 +1602,10 @@ glusterd_mgmt_v3_initiate_all_phases (rpcsvc_request_t *req, glusterd_op_t op,
}
dict_copy (dict, tmp_dict);
- /* BUILD PEERS LIST */
- INIT_LIST_HEAD (&conf->xaction_peers);
- npeers = gd_build_peers_list (&conf->peers, &conf->xaction_peers, op);
-
/* LOCKDOWN PHASE - Acquire mgmt_v3 locks */
- ret = glusterd_mgmt_v3_initiate_lockdown (conf, op, dict, &op_errstr,
- npeers, &is_acquired);
+ ret = glusterd_mgmt_v3_initiate_lockdown (op, dict, &op_errstr,
+ npeers, &is_acquired,
+ &xaction_peers);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "mgmt_v3 lockdown failed.");
goto out;
@@ -1636,16 +1622,17 @@ glusterd_mgmt_v3_initiate_all_phases (rpcsvc_request_t *req, glusterd_op_t op,
}
/* PRE-COMMIT VALIDATE PHASE */
- ret = glusterd_mgmt_v3_pre_validate (conf, op, req_dict,
- &op_errstr, npeers);
+ ret = glusterd_mgmt_v3_pre_validate (op, req_dict,
+ &op_errstr, npeers,
+ &xaction_peers);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "Pre Validation Failed");
goto out;
}
/* COMMIT OP PHASE */
- ret = glusterd_mgmt_v3_commit (conf, op, dict, req_dict,
- &op_errstr, npeers);
+ ret = glusterd_mgmt_v3_commit (op, dict, req_dict,
+ &op_errstr, npeers, &xaction_peers);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "Commit Op Failed");
goto out;
@@ -1656,8 +1643,9 @@ glusterd_mgmt_v3_initiate_all_phases (rpcsvc_request_t *req, glusterd_op_t op,
commands other than snapshot. So as of now, I am
sending 0 (op_ret as 0).
*/
- ret = glusterd_mgmt_v3_post_validate (conf, op, 0, dict, req_dict,
- &op_errstr, npeers);
+ ret = glusterd_mgmt_v3_post_validate (op, 0, dict, req_dict,
+ &op_errstr, npeers,
+ &xaction_peers);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "Post Validation Failed");
goto out;
@@ -1667,9 +1655,10 @@ glusterd_mgmt_v3_initiate_all_phases (rpcsvc_request_t *req, glusterd_op_t op,
out:
op_ret = ret;
/* UNLOCK PHASE FOR PEERS*/
- (void) glusterd_mgmt_v3_release_peer_locks (conf, op, dict,
+ (void) glusterd_mgmt_v3_release_peer_locks (op, dict,
op_ret, &op_errstr,
- npeers, is_acquired);
+ npeers, is_acquired,
+ &xaction_peers);
/* LOCAL VOLUME(S) UNLOCK */
if (is_acquired) {
@@ -1681,10 +1670,12 @@ out:
op_ret = ret;
}
}
-
+rsp:
/* SEND CLI RESPONSE */
glusterd_op_send_cli_response (op, op_ret, 0, req, dict, op_errstr);
+ gd_cleanup_local_xaction_peers_list (&xaction_peers);
+
if (req_dict)
dict_unref (req_dict);
@@ -1777,6 +1768,7 @@ glusterd_mgmt_v3_initiate_snap_phases (rpcsvc_request_t *req, glusterd_op_t op,
uuid_t *originator_uuid = NULL;
gf_boolean_t success = _gf_false;
char *cli_errstr = NULL;
+ struct list_head xaction_peers = {0,};
this = THIS;
GF_ASSERT (this);
@@ -1785,6 +1777,15 @@ glusterd_mgmt_v3_initiate_snap_phases (rpcsvc_request_t *req, glusterd_op_t op,
conf = this->private;
GF_ASSERT (conf);
+ INIT_LIST_HEAD (&xaction_peers);
+ npeers = gd_build_local_xaction_peers_list (&conf->peers,
+ &xaction_peers, op);
+ if (npeers == -1) {
+ gf_log (this->name, GF_LOG_ERROR, "building local peers list "
+ "failed");
+ goto rsp;
+ }
+
/* Save the MY_UUID as the originator_uuid. This originator_uuid
* will be used by is_origin_glusterd() to determine if a node
* is the originator node for a command. */
@@ -1821,13 +1822,10 @@ glusterd_mgmt_v3_initiate_snap_phases (rpcsvc_request_t *req, glusterd_op_t op,
}
dict_copy (dict, tmp_dict);
- /* BUILD PEERS LIST */
- INIT_LIST_HEAD (&conf->xaction_peers);
- npeers = gd_build_peers_list (&conf->peers, &conf->xaction_peers, op);
-
/* LOCKDOWN PHASE - Acquire mgmt_v3 locks */
- ret = glusterd_mgmt_v3_initiate_lockdown (conf, op, dict, &op_errstr,
- npeers, &is_acquired);
+ ret = glusterd_mgmt_v3_initiate_lockdown (op, dict, &op_errstr,
+ npeers, &is_acquired,
+ &xaction_peers);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "mgmt_v3 lockdown failed.");
goto out;
@@ -1844,15 +1842,16 @@ glusterd_mgmt_v3_initiate_snap_phases (rpcsvc_request_t *req, glusterd_op_t op,
}
/* PRE-COMMIT VALIDATE PHASE */
- ret = glusterd_mgmt_v3_pre_validate (conf, op, req_dict,
- &op_errstr, npeers);
+ ret = glusterd_mgmt_v3_pre_validate (op, req_dict,
+ &op_errstr, npeers, &xaction_peers);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "Pre Validation Failed");
goto out;
}
/* quorum check of the volume is done here */
- ret = glusterd_snap_quorum_check (req_dict, _gf_false, &op_errstr);
+ ret = glusterd_snap_quorum_check (req_dict, _gf_false, &op_errstr,
+ &xaction_peers);
if (ret) {
gf_log (this->name, GF_LOG_WARNING,
"Volume quorum check failed");
@@ -1869,8 +1868,8 @@ glusterd_mgmt_v3_initiate_snap_phases (rpcsvc_request_t *req, glusterd_op_t op,
goto out;
}
- ret = glusterd_mgmt_v3_brick_op (conf, op, req_dict,
- &op_errstr, npeers);
+ ret = glusterd_mgmt_v3_brick_op (op, req_dict,
+ &op_errstr, npeers, &xaction_peers);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "Brick Ops Failed");
goto unbarrier;
@@ -1900,8 +1899,8 @@ glusterd_mgmt_v3_initiate_snap_phases (rpcsvc_request_t *req, glusterd_op_t op,
goto unbarrier;
}
- ret = glusterd_mgmt_v3_commit (conf, op, dict, req_dict,
- &op_errstr, npeers);
+ ret = glusterd_mgmt_v3_commit (op, dict, req_dict,
+ &op_errstr, npeers, &xaction_peers);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "Commit Op Failed");
/* If the main op fails, we should save the error string.
@@ -1926,8 +1925,8 @@ unbarrier:
goto out;
}
- ret = glusterd_mgmt_v3_brick_op (conf, op, req_dict,
- &op_errstr, npeers);
+ ret = glusterd_mgmt_v3_brick_op (op, req_dict,
+ &op_errstr, npeers, &xaction_peers);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "Brick Ops Failed");
@@ -1937,7 +1936,8 @@ unbarrier:
/*Do a quorum check if the commit phase is successful*/
if (success) {
//quorum check of the snapshot volume
- ret = glusterd_snap_quorum_check (dict, _gf_true, &op_errstr);
+ ret = glusterd_snap_quorum_check (dict, _gf_true, &op_errstr,
+ &xaction_peers);
if (ret) {
gf_log (this->name, GF_LOG_WARNING,
"Snapshot Volume quorum check failed");
@@ -1954,17 +1954,19 @@ out:
op_ret = -1;
/* POST-COMMIT VALIDATE PHASE */
- ret = glusterd_mgmt_v3_post_validate (conf, op, op_ret, dict, req_dict,
- &op_errstr, npeers);
+ ret = glusterd_mgmt_v3_post_validate (op, op_ret, dict, req_dict,
+ &op_errstr, npeers,
+ &xaction_peers);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "Post Validation Failed");
op_ret = -1;
}
/* UNLOCK PHASE FOR PEERS*/
- (void) glusterd_mgmt_v3_release_peer_locks (conf, op, dict,
+ (void) glusterd_mgmt_v3_release_peer_locks (op, dict,
op_ret, &op_errstr,
- npeers, is_acquired);
+ npeers, is_acquired,
+ &xaction_peers);
/* If the commit op (snapshot taking) failed, then the error is stored
in cli_errstr and unbarrier is called. Suppose, if unbarrier also
@@ -1988,10 +1990,12 @@ out:
op_ret = ret;
}
}
-
+rsp:
/* SEND CLI RESPONSE */
glusterd_op_send_cli_response (op, op_ret, 0, req, dict, op_errstr);
+ gd_cleanup_local_xaction_peers_list (&xaction_peers);
+
if (req_dict)
dict_unref (req_dict);
diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.c b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
index 52ab039b9ec..45fcebffea1 100644
--- a/xlators/mgmt/glusterd/src/glusterd-op-sm.c
+++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
@@ -3457,6 +3457,8 @@ out:
return required;
}
+/* This function shouldn't be used when the quorum validation needs to happen
+ * on the non-global peer list */
static int
glusterd_op_validate_quorum (xlator_t *this, glusterd_op_t op,
dict_t *dict, char **op_errstr)
@@ -3466,7 +3468,6 @@ glusterd_op_validate_quorum (xlator_t *this, glusterd_op_t op,
glusterd_volinfo_t *volinfo = NULL;
char *errstr = NULL;
-
errstr = "Quorum not met. Volume operation not allowed.";
if (!glusterd_is_op_quorum_validation_required (this, op, dict))
goto out;
@@ -3483,7 +3484,10 @@ glusterd_op_validate_quorum (xlator_t *this, glusterd_op_t op,
goto out;
}
- if (does_gd_meet_server_quorum (this, _gf_false)) {
+ /* Passing NULL implies quorum calculation will happen on global peer
+ * list */
+ if (does_gd_meet_server_quorum (this, NULL,
+ _gf_false)) {
ret = 0;
goto out;
}
diff --git a/xlators/mgmt/glusterd/src/glusterd-sm.h b/xlators/mgmt/glusterd/src/glusterd-sm.h
index 294dbacc9d0..fb873f75601 100644
--- a/xlators/mgmt/glusterd/src/glusterd-sm.h
+++ b/xlators/mgmt/glusterd/src/glusterd-sm.h
@@ -103,6 +103,11 @@ struct glusterd_peerinfo_ {
typedef struct glusterd_peerinfo_ glusterd_peerinfo_t;
+typedef struct glusterd_local_peers_ {
+ glusterd_peerinfo_t *peerinfo;
+ struct list_head op_peers_list;
+} glusterd_local_peers_t;
+
typedef enum glusterd_ev_gen_mode_ {
GD_MODE_OFF,
GD_MODE_ON,
diff --git a/xlators/mgmt/glusterd/src/glusterd-syncop.c b/xlators/mgmt/glusterd/src/glusterd-syncop.c
index 8110579de1f..bca8557d3a2 100644
--- a/xlators/mgmt/glusterd/src/glusterd-syncop.c
+++ b/xlators/mgmt/glusterd/src/glusterd-syncop.c
@@ -1027,6 +1027,9 @@ gd_build_peers_list (struct list_head *peers, struct list_head *xact_peers,
glusterd_peerinfo_t *peerinfo = NULL;
int npeers = 0;
+ GF_ASSERT (peers);
+ GF_ASSERT (xact_peers);
+
list_for_each_entry (peerinfo, peers, uuid_list) {
if (!peerinfo->connected)
continue;
@@ -1041,18 +1044,65 @@ gd_build_peers_list (struct list_head *peers, struct list_head *xact_peers,
}
int
-gd_lock_op_phase (glusterd_conf_t *conf, glusterd_op_t op, dict_t *op_ctx,
- char **op_errstr, int npeers, uuid_t txn_id)
+gd_build_local_xaction_peers_list (struct list_head *peers,
+ struct list_head *xact_peers,
+ glusterd_op_t op)
{
- int ret = -1;
- int peer_cnt = 0;
- uuid_t peer_uuid = {0};
- xlator_t *this = NULL;
- glusterd_peerinfo_t *peerinfo = NULL;
- struct syncargs args = {0};
- struct list_head *peers = NULL;
+ glusterd_peerinfo_t *peerinfo = NULL;
+ glusterd_local_peers_t *local_peers = NULL;
+ int npeers = 0;
+
+ GF_ASSERT (peers);
+ GF_ASSERT (xact_peers);
- peers = &conf->xaction_peers;
+ list_for_each_entry (peerinfo, peers, uuid_list) {
+ if (!peerinfo->connected)
+ continue;
+ if (op != GD_OP_SYNC_VOLUME &&
+ peerinfo->state.state != GD_FRIEND_STATE_BEFRIENDED)
+ continue;
+
+ local_peers = GF_CALLOC (1, sizeof (*local_peers),
+ gf_gld_mt_local_peers_t);
+ if (!local_peers) {
+ return -1;
+ }
+ INIT_LIST_HEAD (&local_peers->op_peers_list);
+ local_peers->peerinfo = peerinfo;
+ list_add_tail (&local_peers->op_peers_list, xact_peers);
+ npeers++;
+ }
+ return npeers;
+}
+
+void
+gd_cleanup_local_xaction_peers_list (struct list_head *xact_peers)
+{
+ GF_ASSERT (xact_peers);
+
+ if (list_empty (xact_peers))
+ return;
+
+ glusterd_local_peers_t *local_peers = NULL;
+
+ list_for_each_entry (local_peers, xact_peers, op_peers_list) {
+ GF_FREE (local_peers);
+ /* local_peers->peerinfo need not be freed because it does not
+ * ownership of peerinfo, but merely refer it */
+ }
+}
+
+int
+gd_lock_op_phase (glusterd_conf_t *conf, glusterd_op_t op, dict_t *op_ctx,
+ char **op_errstr, int npeers, uuid_t txn_id,
+ struct list_head *peers)
+{
+ int ret = -1;
+ int peer_cnt = 0;
+ uuid_t peer_uuid = {0};
+ xlator_t *this = NULL;
+ glusterd_peerinfo_t *peerinfo = NULL;
+ struct syncargs args = {0};
if (!npeers) {
ret = 0;
@@ -1062,7 +1112,7 @@ gd_lock_op_phase (glusterd_conf_t *conf, glusterd_op_t op, dict_t *op_ctx,
this = THIS;
synctask_barrier_init((&args));
peer_cnt = 0;
- list_for_each_entry (peerinfo, peers, op_peers_list) {
+ list_for_each_local_xaction_peers (peerinfo, peers) {
if (conf->op_version < GD_OP_VERSION_3_6_0) {
/* Reset lock status */
peerinfo->locked = _gf_false;
@@ -1103,16 +1153,16 @@ int
gd_stage_op_phase (struct list_head *peers, glusterd_op_t op, dict_t *op_ctx,
dict_t *req_dict, char **op_errstr, int npeers)
{
- int ret = -1;
- int peer_cnt = 0;
- dict_t *rsp_dict = NULL;
- char *hostname = NULL;
- xlator_t *this = NULL;
- glusterd_peerinfo_t *peerinfo = NULL;
- uuid_t tmp_uuid = {0};
- char *errstr = NULL;
- struct syncargs args = {0};
- dict_t *aggr_dict = NULL;
+ int ret = -1;
+ int peer_cnt = 0;
+ dict_t *rsp_dict = NULL;
+ char *hostname = NULL;
+ xlator_t *this = NULL;
+ glusterd_peerinfo_t *peerinfo = NULL;
+ uuid_t tmp_uuid = {0};
+ char *errstr = NULL;
+ struct syncargs args = {0};
+ dict_t *aggr_dict = NULL;
this = THIS;
rsp_dict = dict_new ();
@@ -1162,7 +1212,8 @@ stage_done:
gd_syncargs_init (&args, aggr_dict);
synctask_barrier_init((&args));
peer_cnt = 0;
- list_for_each_entry (peerinfo, peers, op_peers_list) {
+
+ list_for_each_local_xaction_peers (peerinfo, peers) {
ret = gd_syncop_mgmt_stage_op (peerinfo->rpc, &args,
MY_UUID, tmp_uuid,
op, req_dict, op_ctx);
@@ -1199,16 +1250,16 @@ int
gd_commit_op_phase (struct list_head *peers, glusterd_op_t op, dict_t *op_ctx,
dict_t *req_dict, char **op_errstr, int npeers)
{
- dict_t *rsp_dict = NULL;
- int peer_cnt = -1;
- int ret = -1;
- char *hostname = NULL;
- glusterd_peerinfo_t *peerinfo = NULL;
- xlator_t *this = NULL;
- uuid_t tmp_uuid = {0};
- char *errstr = NULL;
- struct syncargs args = {0};
- int type = GF_QUOTA_OPTION_TYPE_NONE;
+ dict_t *rsp_dict = NULL;
+ int peer_cnt = -1;
+ int ret = -1;
+ char *hostname = NULL;
+ glusterd_peerinfo_t *peerinfo = NULL;
+ xlator_t *this = NULL;
+ uuid_t tmp_uuid = {0};
+ char *errstr = NULL;
+ struct syncargs args = {0};
+ int type = GF_QUOTA_OPTION_TYPE_NONE;
this = THIS;
rsp_dict = dict_new ();
@@ -1266,7 +1317,8 @@ commit_done:
gd_syncargs_init (&args, op_ctx);
synctask_barrier_init((&args));
peer_cnt = 0;
- list_for_each_entry (peerinfo, peers, op_peers_list) {
+
+ list_for_each_local_xaction_peers (peerinfo, peers) {
ret = gd_syncop_mgmt_commit_op (peerinfo->rpc, &args,
MY_UUID, tmp_uuid,
op, req_dict, op_ctx);
@@ -1298,18 +1350,15 @@ int
gd_unlock_op_phase (glusterd_conf_t *conf, glusterd_op_t op, int *op_ret,
rpcsvc_request_t *req, dict_t *op_ctx, char *op_errstr,
int npeers, char *volname, gf_boolean_t is_acquired,
- uuid_t txn_id)
+ uuid_t txn_id, struct list_head *peers)
{
- glusterd_peerinfo_t *peerinfo = NULL;
- glusterd_peerinfo_t *tmp = NULL;
- uuid_t tmp_uuid = {0};
- int peer_cnt = 0;
- int ret = -1;
- xlator_t *this = NULL;
- struct syncargs args = {0};
- struct list_head *peers = NULL;
-
- peers = &conf->xaction_peers;
+ glusterd_peerinfo_t *peerinfo = NULL;
+ uuid_t tmp_uuid = {0};
+ int peer_cnt = 0;
+ int ret = -1;
+ xlator_t *this = NULL;
+ struct syncargs args = {0};
+
this = THIS;
GF_ASSERT (this);
@@ -1327,25 +1376,23 @@ gd_unlock_op_phase (glusterd_conf_t *conf, glusterd_op_t op, int *op_ret,
synctask_barrier_init((&args));
peer_cnt = 0;
+
if (conf->op_version < GD_OP_VERSION_3_6_0) {
- list_for_each_entry_safe (peerinfo, tmp, peers, op_peers_list) {
+ list_for_each_local_xaction_peers (peerinfo, peers) {
/* Only unlock peers that were locked */
if (peerinfo->locked) {
gd_syncop_mgmt_unlock (peerinfo, &args,
MY_UUID, tmp_uuid);
peer_cnt++;
- list_del_init (&peerinfo->op_peers_list);
}
}
} else {
if (volname) {
- list_for_each_entry_safe (peerinfo, tmp,
- peers, op_peers_list) {
+ list_for_each_local_xaction_peers (peerinfo, peers) {
gd_syncop_mgmt_v3_unlock (op_ctx, peerinfo,
&args, MY_UUID,
tmp_uuid, txn_id);
peer_cnt++;
- list_del_init (&peerinfo->op_peers_list);
}
}
}
@@ -1499,6 +1546,7 @@ gd_sync_task_begin (dict_t *op_ctx, rpcsvc_request_t * req)
xlator_t *this = NULL;
gf_boolean_t is_acquired = _gf_false;
uuid_t *txn_id = NULL;
+ struct list_head xaction_peers = {0,};
glusterd_op_info_t txn_opinfo;
this = THIS;
@@ -1506,6 +1554,8 @@ gd_sync_task_begin (dict_t *op_ctx, rpcsvc_request_t * req)
conf = this->private;
GF_ASSERT (conf);
+ INIT_LIST_HEAD (&xaction_peers);
+
ret = dict_get_int32 (op_ctx, GD_SYNC_OPCODE_KEY, &tmp_op);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "Failed to get volume "
@@ -1588,15 +1638,20 @@ gd_sync_task_begin (dict_t *op_ctx, rpcsvc_request_t * req)
local_locking_done:
- INIT_LIST_HEAD (&conf->xaction_peers);
-
- npeers = gd_build_peers_list (&conf->peers, &conf->xaction_peers, op);
+ /* Maintain xaction_peers on per transaction basis */
+ npeers = gd_build_local_xaction_peers_list (&conf->peers,
+ &xaction_peers, op);
+ if (npeers == -1) {
+ gf_log (this->name, GF_LOG_ERROR, "building local peers list "
+ "failed");
+ goto out;
+ }
/* If no volname is given as a part of the command, locks will
* not be held */
if (volname || (conf->op_version < GD_OP_VERSION_3_6_0)) {
ret = gd_lock_op_phase (conf, op, op_ctx, &op_errstr,
- npeers, *txn_id);
+ npeers, *txn_id, &xaction_peers);
if (ret) {
gf_log (this->name, GF_LOG_ERROR,
"Locking Peers Failed.");
@@ -1613,7 +1668,7 @@ local_locking_done:
goto out;
}
- ret = gd_stage_op_phase (&conf->xaction_peers, op, op_ctx, req_dict,
+ ret = gd_stage_op_phase (&xaction_peers, op, op_ctx, req_dict,
&op_errstr, npeers);
if (ret)
goto out;
@@ -1622,7 +1677,7 @@ local_locking_done:
if (ret)
goto out;
- ret = gd_commit_op_phase (&conf->xaction_peers, op, op_ctx, req_dict,
+ ret = gd_commit_op_phase (&xaction_peers, op, op_ctx, req_dict,
&op_errstr, npeers);
if (ret)
goto out;
@@ -1634,7 +1689,8 @@ out:
(void) gd_unlock_op_phase (conf, op, &op_ret, req,
op_ctx, op_errstr,
npeers, volname,
- is_acquired, *txn_id);
+ is_acquired, *txn_id,
+ &xaction_peers);
/* Clearing the transaction opinfo */
ret = glusterd_clear_txn_opinfo (txn_id);
@@ -1647,6 +1703,8 @@ out:
glusterd_op_send_cli_response (op, op_ret, 0, req, op_ctx, op_errstr);
+ gd_cleanup_local_xaction_peers_list (&xaction_peers);
+
if (volname)
GF_FREE (volname);
diff --git a/xlators/mgmt/glusterd/src/glusterd-syncop.h b/xlators/mgmt/glusterd/src/glusterd-syncop.h
index e83ea2f4c46..1944bb4e143 100644
--- a/xlators/mgmt/glusterd/src/glusterd-syncop.h
+++ b/xlators/mgmt/glusterd/src/glusterd-syncop.h
@@ -37,11 +37,19 @@
} while (0)
+#define list_for_each_local_xaction_peers(xact_peer, xact_peers_head) \
+ glusterd_local_peers_t *pos = NULL; \
+ for (pos = list_entry((xact_peers_head)->next, glusterd_local_peers_t,\
+ op_peers_list), \
+ xact_peer = pos->peerinfo; \
+ &pos->op_peers_list != (xact_peers_head); \
+ pos = list_entry(pos->op_peers_list.next, glusterd_local_peers_t,\
+ op_peers_list), \
+ xact_peer = pos->peerinfo)
+
int gd_syncop_submit_request (struct rpc_clnt *rpc, void *req, void *local,
void *cookie, rpc_clnt_prog_t *prog, int procnum,
fop_cbk_fn_t cbkfn, xdrproc_t xdrproc);
-
-
int gd_syncop_mgmt_lock (glusterd_peerinfo_t *peerinfo, struct syncargs *arg,
uuid_t my_uuid, uuid_t recv_uuid);
int gd_syncop_mgmt_unlock (glusterd_peerinfo_t *peerinfo, struct syncargs *arg,
@@ -59,6 +67,15 @@ gd_synctask_barrier_wait (struct syncargs *args, int count);
int
gd_build_peers_list (struct list_head *peers, struct list_head *xact_peers,
glusterd_op_t op);
+
+int
+gd_build_local_xaction_peers_list (struct list_head *peers,
+ struct list_head *xact_peers,
+ glusterd_op_t op);
+
+void
+gd_cleanup_local_xaction_peers_list (struct list_head *peers);
+
int
gd_brick_op_phase (glusterd_op_t op, dict_t *op_ctx, dict_t *req_dict,
char **op_errstr);
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index dba22fd0113..ef53128e8c4 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -3489,7 +3489,8 @@ _does_quorum_meet (int active_count, int quorum_count)
int
glusterd_get_quorum_cluster_counts (xlator_t *this, int *active_count,
int *quorum_count,
- gf_boolean_t _xaction_peers)
+ struct list_head *peer_list,
+ gf_boolean_t _local_xaction_peers)
{
glusterd_peerinfo_t *peerinfo = NULL;
glusterd_conf_t *conf = NULL;
@@ -3499,25 +3500,34 @@ glusterd_get_quorum_cluster_counts (xlator_t *this, int *active_count,
double quorum_percentage = 0.0;
gf_boolean_t ratio = _gf_false;
int count = 0;
- struct list_head *peer_list = NULL;
conf = this->private;
+
+ GF_ASSERT (conf);
+
//Start with counting self
inquorum_count = 1;
if (active_count)
*active_count = 1;
- peer_list = (_xaction_peers) ? &conf->xaction_peers : &conf->peers;
-
- if (_xaction_peers) {
- list_for_each_entry (peerinfo, peer_list, op_peers_list) {
+ if (!peer_list) {
+ list_for_each_entry (peerinfo, &conf->peers, uuid_list) {
glusterd_quorum_count(peerinfo, inquorum_count,
active_count, out);
}
} else {
- list_for_each_entry (peerinfo, peer_list, uuid_list) {
- glusterd_quorum_count(peerinfo, inquorum_count,
- active_count, out);
+ if (_local_xaction_peers) {
+ list_for_each_local_xaction_peers (peerinfo,
+ peer_list) {
+ glusterd_quorum_count(peerinfo, inquorum_count,
+ active_count, out);
+ }
+ } else {
+ list_for_each_entry (peerinfo, peer_list,
+ op_peers_list) {
+ glusterd_quorum_count(peerinfo, inquorum_count,
+ active_count, out);
+ }
}
}
@@ -3574,7 +3584,8 @@ glusterd_is_any_volume_in_server_quorum (xlator_t *this)
}
gf_boolean_t
-does_gd_meet_server_quorum (xlator_t *this, gf_boolean_t _xaction_peers)
+does_gd_meet_server_quorum (xlator_t *this, struct list_head *peers_list,
+ gf_boolean_t _local_xaction_peers)
{
int quorum_count = 0;
int active_count = 0;
@@ -3585,7 +3596,8 @@ does_gd_meet_server_quorum (xlator_t *this, gf_boolean_t _xaction_peers)
conf = this->private;
ret = glusterd_get_quorum_cluster_counts (this, &active_count,
&quorum_count,
- _xaction_peers);
+ peers_list,
+ _local_xaction_peers);
if (ret)
goto out;
@@ -3701,7 +3713,7 @@ glusterd_do_quorum_action ()
{
ret = glusterd_get_quorum_cluster_counts (this, &active_count,
- &quorum_count,
+ &quorum_count, NULL,
_gf_false);
if (ret)
goto unlock;
@@ -12795,7 +12807,8 @@ out:
int32_t
glusterd_snap_quorum_check_for_create (dict_t *dict, gf_boolean_t snap_volume,
- char **op_errstr)
+ char **op_errstr,
+ struct list_head *peers_list)
{
int8_t snap_force = 0;
int32_t force = 0;
@@ -12848,7 +12861,7 @@ glusterd_snap_quorum_check_for_create (dict_t *dict, gf_boolean_t snap_volume,
by glusterd and if glusterds are not in
quorum, then better fail the snapshot
*/
- if (!does_gd_meet_server_quorum (this,_gf_true)) {
+ if (!does_gd_meet_server_quorum (this, peers_list, _gf_true)) {
snprintf (err_str, sizeof (err_str),
"glusterds are not in quorum");
gf_log (this->name, GF_LOG_WARNING, "%s",
@@ -12989,7 +13002,8 @@ out:
int32_t
glusterd_snap_quorum_check (dict_t *dict, gf_boolean_t snap_volume,
- char **op_errstr)
+ char **op_errstr,
+ struct list_head *peers_list)
{
int32_t ret = -1;
xlator_t *this = NULL;
@@ -13015,7 +13029,8 @@ glusterd_snap_quorum_check (dict_t *dict, gf_boolean_t snap_volume,
switch (snap_command) {
case GF_SNAP_OPTION_TYPE_CREATE:
ret = glusterd_snap_quorum_check_for_create (dict, snap_volume,
- op_errstr);
+ op_errstr,
+ peers_list);
if (ret) {
gf_log (this->name, GF_LOG_WARNING, "Quorum check"
"failed during snapshot create command");
@@ -13024,7 +13039,7 @@ glusterd_snap_quorum_check (dict_t *dict, gf_boolean_t snap_volume,
break;
case GF_SNAP_OPTION_TYPE_DELETE:
case GF_SNAP_OPTION_TYPE_RESTORE:
- if (!does_gd_meet_server_quorum (this, _gf_true)) {
+ if (!does_gd_meet_server_quorum (this, peers_list, _gf_true)) {
ret = -1;
snprintf (err_str, sizeof (err_str),
"glusterds are not in quorum");
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.h b/xlators/mgmt/glusterd/src/glusterd-utils.h
index 368b6627159..636f72613d9 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.h
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.h
@@ -566,7 +566,8 @@ glusterd_do_quorum_action ();
int
glusterd_get_quorum_cluster_counts (xlator_t *this, int *active_count,
int *quorum_count,
- gf_boolean_t _xaction_peers);
+ struct list_head *peer_list,
+ gf_boolean_t _local__xaction_peers);
int
glusterd_get_next_global_opt_version_str (dict_t *opts, char **version_str);
@@ -577,7 +578,9 @@ glusterd_is_volume_in_server_quorum (glusterd_volinfo_t *volinfo);
gf_boolean_t
glusterd_is_any_volume_in_server_quorum (xlator_t *this);
gf_boolean_t
-does_gd_meet_server_quorum (xlator_t *this, gf_boolean_t _xaction_peers);
+does_gd_meet_server_quorum (xlator_t *this,
+ struct list_head *peers_list,
+ gf_boolean_t _local__xaction_peers);
int
glusterd_generate_and_set_task_id (dict_t *dict, char *key);
@@ -785,11 +788,13 @@ glusterd_take_lvm_snapshot (glusterd_brickinfo_t *brickinfo,
int32_t
glusterd_snap_quorum_check (dict_t *dict, gf_boolean_t snap_volume,
- char **op_errstr);
+ char **op_errstr,
+ struct list_head *peers_list);
int32_t
glusterd_snap_quorum_check_for_create (dict_t *dict, gf_boolean_t snap_volume,
- char **op_errstr);
+ char **op_errstr,
+ struct list_head *peers_list);
int32_t
glusterd_volume_quorum_check (glusterd_volinfo_t *volinfo, int64_t index,