diff options
Diffstat (limited to 'xlators/mgmt/glusterd/src/glusterd-op-sm.c')
| -rw-r--r-- | xlators/mgmt/glusterd/src/glusterd-op-sm.c | 1298 |
1 files changed, 1049 insertions, 249 deletions
diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.c b/xlators/mgmt/glusterd/src/glusterd-op-sm.c index 4e4f0394d..1666f5e4d 100644 --- a/xlators/mgmt/glusterd/src/glusterd-op-sm.c +++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.c @@ -37,6 +37,7 @@ #include "glusterd-store.h" #include "glusterd-hooks.h" #include "glusterd-volgen.h" +#include "glusterd-locks.h" #include "syscall.h" #include "cli1-xdr.h" #include "common-utils.h" @@ -67,6 +68,165 @@ static struct list_head gd_op_sm_queue; pthread_mutex_t gd_op_sm_lock; glusterd_op_info_t opinfo = {{0},}; +uuid_t global_txn_id = {"\0"}; /* To be used in + * heterogeneous + * cluster with no + * transaction ids */ + +static dict_t *txn_opinfo; + +struct txn_opinfo_object_ { + glusterd_op_info_t opinfo; +}; +typedef struct txn_opinfo_object_ txn_opinfo_obj; + +int32_t +glusterd_txn_opinfo_dict_init () +{ + int32_t ret = -1; + + txn_opinfo = dict_new (); + if (!txn_opinfo) { + ret = -1; + goto out; + } + + ret = 0; +out: + return ret; +} + +void +glusterd_txn_opinfo_dict_fini () +{ + if (txn_opinfo) + dict_destroy (txn_opinfo); +} + +void +glusterd_txn_opinfo_init (glusterd_op_info_t *opinfo, + glusterd_op_sm_state_info_t *state, + glusterd_op_t *op, + dict_t *op_ctx, + rpcsvc_request_t *req) +{ + GF_ASSERT (opinfo); + + if (state) + opinfo->state = *state; + + if (op) + opinfo->op = *op; + + opinfo->op_ctx = dict_ref(op_ctx); + + if (req) + opinfo->req = req; + + return; +} + +int32_t +glusterd_get_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t *opinfo) +{ + int32_t ret = -1; + txn_opinfo_obj *opinfo_obj = NULL; + + if (!txn_id || !opinfo) { + gf_log ("", GF_LOG_ERROR, + "Empty transaction id or opinfo received."); + ret = -1; + goto out; + } + + ret = dict_get_bin(txn_opinfo, uuid_utoa (*txn_id), + (void **) &opinfo_obj); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to get transaction opinfo"); + goto out; + } + + (*opinfo) = opinfo_obj->opinfo; + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_set_txn_opinfo (uuid_t *txn_id, glusterd_op_info_t *opinfo) +{ + int32_t ret = -1; + txn_opinfo_obj *opinfo_obj = NULL; + + if (!txn_id) { + gf_log ("", GF_LOG_ERROR, "Empty transaction id received."); + ret = -1; + goto out; + } + + ret = dict_get_bin(txn_opinfo, uuid_utoa (*txn_id), + (void **) &opinfo_obj); + if (ret) { + opinfo_obj = GF_CALLOC (1, sizeof(txn_opinfo_obj), + gf_common_mt_txn_opinfo_obj_t); + if (!opinfo_obj) { + ret = -1; + goto out; + } + + ret = dict_set_bin(txn_opinfo, uuid_utoa (*txn_id), opinfo_obj, + sizeof(txn_opinfo_obj)); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Unable to set opinfo for transaction ID : %s", + uuid_utoa (*txn_id)); + goto out; + } + } + + opinfo_obj->opinfo = (*opinfo); + + ret = 0; +out: + if (ret) + if (opinfo_obj) + GF_FREE (opinfo_obj); + + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + +int32_t +glusterd_clear_txn_opinfo (uuid_t *txn_id) +{ + int32_t ret = -1; + glusterd_op_info_t txn_op_info = {{0},}; + + if (!txn_id) { + gf_log ("", GF_LOG_ERROR, "Empty transaction id received."); + ret = -1; + goto out; + } + + ret = glusterd_get_txn_opinfo (txn_id, &txn_op_info); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Transaction opinfo not found"); + goto out; + } + + dict_unref (txn_op_info.op_ctx); + + dict_del(txn_opinfo, uuid_utoa (*txn_id)); + + ret = 0; +out: + gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + return ret; +} + static int glusterfs_port = GLUSTERD_DEFAULT_PORT; static char *glusterd_op_sm_state_names[] = { "Default", @@ -147,10 +307,10 @@ glusterd_is_volume_started (glusterd_volinfo_t *volinfo) } static int -glusterd_op_sm_inject_all_acc () +glusterd_op_sm_inject_all_acc (uuid_t *txn_id) { int32_t ret = -1; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, txn_id, NULL); gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -235,20 +395,20 @@ glusterd_brick_op_build_payload (glusterd_op_t op, glusterd_brickinfo_t *brickin brick_req->name = gf_strdup (name); break; - -#ifdef HAVE_BD_XLATOR - case GD_OP_BD_OP: - { + case GD_OP_SNAP: brick_req = GF_CALLOC (1, sizeof (*brick_req), gf_gld_mt_mop_brick_req_t); if (!brick_req) goto out; - brick_req->op = GLUSTERD_BRICK_BD_OP; - brick_req->name = ""; - } + brick_req->op = GLUSTERD_VOLUME_BARRIER_OP; + ret = dict_get_str (dict, "volname", &volname); + if (ret) + goto out; + snprintf (name, 1024, "%s-server",volname); + brick_req->name = gf_strdup (name); + break; -#endif default: goto out; break; @@ -416,11 +576,17 @@ glusterd_op_stage_set_volume (dict_t *dict, char **op_errstr) if (!val_dict) goto out; + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, "Unable to get volume name"); + goto out; + } + /* Check if we can support the required op-version * This check is not done on the originator glusterd. The originator * glusterd sets this value. */ - origin_glusterd = is_origin_glusterd (); + origin_glusterd = is_origin_glusterd (dict); if (!origin_glusterd) { /* Check for v3.3.x origin glusterd */ @@ -828,7 +994,7 @@ glusterd_op_stage_sync_volume (dict_t *dict, char **op_errstr) goto out; } - if (glusterd_is_local_addr (hostname)) { + if (gf_is_local_addr (hostname)) { //volname is not present in case of sync all ret = dict_get_str (dict, "volname", &volname); if (!ret) { @@ -1101,14 +1267,17 @@ _delete_reconfig_opt (dict_t *this, char *key, data_t *value, void *data) GF_ASSERT (data); is_force = (int32_t*)data; - if (*is_force != 1 && - (_gf_true == glusterd_check_voloption_flags (key, - OPT_FLAG_FORCE))) { + if (*is_force != 1) { + if (_gf_true == glusterd_check_voloption_flags (key, + OPT_FLAG_FORCE)) { /* indicate to caller that we don't set the option * due to being protected */ - *is_force = -1; - goto out; + *is_force = *is_force | GD_OP_PROTECTED; + goto out; + } else { + *is_force = *is_force | GD_OP_UNPROTECTED; + } } gf_log ("", GF_LOG_DEBUG, "deleting dict with key=%s,value=%s", @@ -1160,8 +1329,9 @@ glusterd_options_reset (glusterd_volinfo_t *volinfo, char *key, _delete_reconfig_opt (volinfo->dict, key, value, is_force); } - ret = glusterd_create_volfiles_and_notify_services (volinfo); + gd_update_volume_op_versions (volinfo); + ret = glusterd_create_volfiles_and_notify_services (volinfo); if (ret) { gf_log (this->name, GF_LOG_ERROR, "Unable to create volfile for" " 'volume reset'"); @@ -1276,7 +1446,7 @@ out: } static int -glusterd_op_reset_volume (dict_t *dict, char **op_errstr) +glusterd_op_reset_volume (dict_t *dict, char **op_rspstr) { glusterd_volinfo_t *volinfo = NULL; int ret = -1; @@ -1331,10 +1501,18 @@ glusterd_op_reset_volume (dict_t *dict, char **op_errstr) quorum_action = _gf_true; ret = glusterd_options_reset (volinfo, key, &is_force); - if (is_force == -1) { - ret = -1; - gf_asprintf(op_errstr, "'%s' is protected. To reset use 'force'.", - key); + if (ret == -1) { + gf_asprintf(op_rspstr, "Volume reset : failed"); + } else if (is_force & GD_OP_PROTECTED) { + if (is_force & GD_OP_UNPROTECTED) { + gf_asprintf (op_rspstr, "All unprotected fields were" + " reset. To reset the protected fields," + " use 'force'."); + } else { + ret = -1; + gf_asprintf (op_rspstr, "'%s' is protected. To reset" + " use 'force'.", key); + } } out: @@ -1364,14 +1542,25 @@ glusterd_stop_bricks (glusterd_volinfo_t *volinfo) int glusterd_start_bricks (glusterd_volinfo_t *volinfo) { - glusterd_brickinfo_t *brickinfo = NULL; + int ret = -1; + glusterd_brickinfo_t *brickinfo = NULL; + + GF_ASSERT (volinfo); list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { - if (glusterd_brick_start (volinfo, brickinfo, _gf_false)) - return -1; + ret = glusterd_brick_start (volinfo, brickinfo, _gf_false); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "Failed to start %s:%s for %s", + brickinfo->hostname, brickinfo->path, + volinfo->volname); + goto out; + } } - return 0; + ret = 0; +out: + return ret; } static int @@ -1474,6 +1663,7 @@ glusterd_op_set_volume (dict_t *dict) char str[50] = {0, }; char *op_errstr = NULL; gf_boolean_t global_opt = _gf_false; + gf_boolean_t global_opts_set = _gf_false; glusterd_volinfo_t *voliter = NULL; int32_t dict_count = 0; gf_boolean_t check_op_version = _gf_false; @@ -1520,6 +1710,7 @@ glusterd_op_set_volume (dict_t *dict) goto out; } + // TODO: Remove this once v3.3 compatability is not required check_op_version = dict_get_str_boolean (dict, "check-op-version", _gf_false); @@ -1534,7 +1725,6 @@ glusterd_op_set_volume (dict_t *dict) for (count = 1; ret != -1 ; count++) { - global_opt = _gf_false; sprintf (str, "key%d", count); ret = dict_get_str (dict, str, &key); if (ret) @@ -1582,8 +1772,11 @@ glusterd_op_set_volume (dict_t *dict) } } - if (glusterd_check_globaloption (key)) + global_opt = _gf_false; + if (glusterd_check_globaloption (key)) { global_opt = _gf_true; + global_opts_set = _gf_true; + } if (!global_opt) value = gf_strdup (value); @@ -1626,7 +1819,21 @@ glusterd_op_set_volume (dict_t *dict) goto out; } - if (!global_opt) { + /* Update the cluster op-version before regenerating volfiles so that + * correct volfiles are generated + */ + if (new_op_version > priv->op_version) { + priv->op_version = new_op_version; + ret = glusterd_store_global_info (this); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to store op-version"); + goto out; + } + } + + if (!global_opts_set) { + gd_update_volume_op_versions (volinfo); ret = glusterd_create_volfiles_and_notify_services (volinfo); if (ret) { gf_log (this->name, GF_LOG_ERROR, @@ -1652,6 +1859,7 @@ glusterd_op_set_volume (dict_t *dict) } else { list_for_each_entry (voliter, &priv->volumes, vol_list) { volinfo = voliter; + gd_update_volume_op_versions (volinfo); ret = glusterd_create_volfiles_and_notify_services (volinfo); if (ret) { gf_log (this->name, GF_LOG_ERROR, @@ -1677,16 +1885,6 @@ glusterd_op_set_volume (dict_t *dict) } } - if (new_op_version > priv->op_version) { - priv->op_version = new_op_version; - ret = glusterd_store_global_info (this); - if (ret) { - gf_log (this->name, GF_LOG_ERROR, - "Failed to store op-version"); - goto out; - } - } - out: GF_FREE (key_fixed); gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret); @@ -1723,7 +1921,7 @@ glusterd_op_sync_volume (dict_t *dict, char **op_errstr, goto out; } - if (!glusterd_is_local_addr (hostname)) { + if (!gf_is_local_addr (hostname)) { ret = 0; goto out; } @@ -1894,6 +2092,105 @@ out: } static int +_add_brick_name_to_dict (dict_t *dict, char *key, glusterd_brickinfo_t *brick) +{ + int ret = -1; + char tmp[1024] = {0,}; + char *brickname = NULL; + xlator_t *this = NULL; + + GF_ASSERT (dict); + GF_ASSERT (key); + GF_ASSERT (brick); + + this = THIS; + GF_ASSERT (this); + + snprintf (tmp, sizeof (tmp), "%s:%s", brick->hostname, brick->path); + brickname = gf_strdup (tmp); + if (!brickname) { + gf_log (this->name, GF_LOG_ERROR, "Failed to dup brick name"); + goto out; + } + + ret = dict_set_dynstr (dict, key, brickname); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add brick name to dict"); + goto out; + } + brickname = NULL; +out: + if (brickname) + GF_FREE (brickname); + return ret; +} + +static int +_add_remove_bricks_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo, + char *prefix) +{ + int ret = -1; + int count = 0; + int i = 0; + char brick_key[1024] = {0,}; + char dict_key[1024] ={0,}; + char *brick = NULL; + xlator_t *this = NULL; + + GF_ASSERT (dict); + GF_ASSERT (volinfo); + GF_ASSERT (prefix); + + this = THIS; + GF_ASSERT (this); + + ret = dict_get_int32 (volinfo->rebal.dict, "count", &count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to get brick count"); + goto out; + } + + snprintf (dict_key, sizeof (dict_key), "%s.count", prefix); + ret = dict_set_int32 (dict, dict_key, count); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to set brick count in dict"); + goto out; + } + + for (i = 1; i <= count; i++) { + memset (brick_key, 0, sizeof (brick_key)); + snprintf (brick_key, sizeof (brick_key), "brick%d", i); + + ret = dict_get_str (volinfo->rebal.dict, brick_key, &brick); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get %s", brick_key); + goto out; + } + + memset (dict_key, 0, sizeof (dict_key)); + snprintf (dict_key, sizeof (dict_key), "%s.%s", prefix, + brick_key); + ret = dict_set_str (dict, dict_key, brick); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add brick to dict"); + goto out; + } + brick = NULL; + } + +out: + return ret; +} + +/* This adds the respective task-id and all available parameters of a task into + * a dictionary + */ +static int _add_task_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo, int op, int index) { @@ -1910,13 +2207,34 @@ _add_task_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo, int op, int index) GF_ASSERT (this); switch (op) { - case GD_OP_REBALANCE: case GD_OP_REMOVE_BRICK: + snprintf (key, sizeof (key), "task%d", index); + ret = _add_remove_bricks_to_dict (dict, volinfo, key); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add remove bricks to dict"); + goto out; + } + case GD_OP_REBALANCE: uuid_str = gf_strdup (uuid_utoa (volinfo->rebal.rebalance_id)); status = volinfo->rebal.defrag_status; break; case GD_OP_REPLACE_BRICK: + snprintf (key, sizeof (key), "task%d.src-brick", index); + ret = _add_brick_name_to_dict (dict, key, + volinfo->rep_brick.src_brick); + if (ret) + goto out; + memset (key, 0, sizeof (key)); + + snprintf (key, sizeof (key), "task%d.dst-brick", index); + ret = _add_brick_name_to_dict (dict, key, + volinfo->rep_brick.dst_brick); + if (ret) + goto out; + memset (key, 0, sizeof (key)); + uuid_str = gf_strdup (uuid_utoa (volinfo->rep_brick.rb_id)); status = volinfo->rep_brick.rb_status; break; @@ -1929,8 +2247,7 @@ _add_task_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo, int op, int index) } snprintf (key, sizeof (key), "task%d.type", index); - ret = dict_set_str (dict, key, - (char *)gd_op_list[op]); + ret = dict_set_str (dict, key, (char *)gd_op_list[op]); if (ret) { gf_log (this->name, GF_LOG_ERROR, "Error setting task type in dict"); @@ -1940,7 +2257,6 @@ _add_task_to_dict (dict_t *dict, glusterd_volinfo_t *volinfo, int op, int index) memset (key, 0, sizeof (key)); snprintf (key, sizeof (key), "task%d.id", index); - if (!uuid_str) goto out; ret = dict_set_dynstr (dict, key, uuid_str); @@ -1967,6 +2283,50 @@ out: } static int +glusterd_aggregate_task_status (dict_t *rsp_dict, glusterd_volinfo_t *volinfo) +{ + int ret = -1; + int tasks = 0; + xlator_t *this = NULL; + + this = THIS; + GF_ASSERT (this); + + if (!uuid_is_null (volinfo->rebal.rebalance_id)) { + ret = _add_task_to_dict (rsp_dict, volinfo, volinfo->rebal.op, + tasks); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add task details to dict"); + goto out; + } + tasks++; + } + + if (!uuid_is_null (volinfo->rep_brick.rb_id)) { + ret = _add_task_to_dict (rsp_dict, volinfo, GD_OP_REPLACE_BRICK, + tasks); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Failed to add task details to dict"); + goto out; + } + tasks++; + } + + ret = dict_set_int32 (rsp_dict, "tasks", tasks); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Error setting tasks count in dict"); + goto out; + } + ret = 0; + +out: + return ret; +} + +static int glusterd_op_status_volume (dict_t *dict, char **op_errstr, dict_t *rsp_dict) { @@ -1986,7 +2346,6 @@ glusterd_op_status_volume (dict_t *dict, char **op_errstr, gf_boolean_t nfs_disabled = _gf_false; gf_boolean_t shd_enabled = _gf_true; gf_boolean_t origin_glusterd = _gf_false; - int tasks = 0; this = THIS; GF_ASSERT (this); @@ -1996,13 +2355,13 @@ glusterd_op_status_volume (dict_t *dict, char **op_errstr, GF_ASSERT (dict); - origin_glusterd = is_origin_glusterd (); + origin_glusterd = is_origin_glusterd (dict); ret = dict_get_uint32 (dict, "cmd", &cmd); if (ret) goto out; - if (is_origin_glusterd ()) { + if (origin_glusterd) { ret = 0; if ((cmd & GF_CLI_STATUS_ALL)) { ret = glusterd_get_all_volnames (rsp_dict); @@ -2069,6 +2428,10 @@ glusterd_op_status_volume (dict_t *dict, char **op_errstr, brick_index); node_count++; + } else if ((cmd & GF_CLI_STATUS_TASKS) != 0) { + ret = glusterd_aggregate_task_status (rsp_dict, volinfo); + goto out; + } else { list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { brick_index++; @@ -2142,35 +2505,18 @@ glusterd_op_status_volume (dict_t *dict, char **op_errstr, } /* Active tasks */ + /* Tasks are added only for normal volume status request for either a + * single volume or all volumes, and only by the origin glusterd + */ if (((cmd & GF_CLI_STATUS_MASK) != GF_CLI_STATUS_NONE) || + !(cmd & (GF_CLI_STATUS_VOL | GF_CLI_STATUS_ALL)) || !origin_glusterd) goto out; - if (!uuid_is_null (volinfo->rebal.rebalance_id)) { - ret = _add_task_to_dict (rsp_dict, volinfo, volinfo->rebal.op, - tasks); - if (ret) { - gf_log (this->name, GF_LOG_ERROR, - "Failed to add task details to dict"); - goto out; - } - tasks++; - } - if (!uuid_is_null (volinfo->rep_brick.rb_id)) { - ret = _add_task_to_dict (rsp_dict, volinfo, GD_OP_REPLACE_BRICK, - tasks); - if (ret) { - gf_log (this->name, GF_LOG_ERROR, - "Failed to add task details to dict"); - goto out; - } - tasks++; - } - - ret = dict_set_int32 (rsp_dict, "tasks", tasks); + ret = glusterd_aggregate_task_status (rsp_dict, volinfo); if (ret) - gf_log (this->name, GF_LOG_ERROR, - "Error setting tasks count in dict"); + goto out; + ret = 0; out: gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -2197,6 +2543,7 @@ glusterd_op_ac_send_lock (glusterd_op_sm_event_t *event, void *ctx) xlator_t *this = NULL; glusterd_peerinfo_t *peerinfo = NULL; uint32_t pending_count = 0; + dict_t *dict = NULL; this = THIS; priv = this->private; @@ -2211,27 +2558,60 @@ glusterd_op_ac_send_lock (glusterd_op_sm_event_t *event, void *ctx) (glusterd_op_get_op() != GD_OP_SYNC_VOLUME)) continue; - proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_LOCK]; - if (proc->fn) { - ret = proc->fn (NULL, this, peerinfo); - if (ret) { - gf_log (this->name, GF_LOG_WARNING, "Failed to " - "send lock request for operation " - "'Volume %s' to peer %s", - gd_op_list[opinfo.op], - peerinfo->hostname); - continue; + /* Based on the op_version, acquire a cluster or mgmt_v3 lock */ + if (priv->op_version < 3) { + proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_LOCK]; + if (proc->fn) { + ret = proc->fn (NULL, this, peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Failed to send lock request " + "for operation 'Volume %s' to " + "peer %s", + gd_op_list[opinfo.op], + peerinfo->hostname); + continue; + } + pending_count++; + } + } else { + dict = glusterd_op_get_ctx (); + dict_ref (dict); + + proc = &peerinfo->mgmt_v3->proctable + [GLUSTERD_MGMT_V3_LOCK]; + if (proc->fn) { + ret = dict_set_static_ptr (dict, "peerinfo", + peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "failed to set peerinfo"); + dict_unref (dict); + goto out; + } + + ret = proc->fn (NULL, this, dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Failed to send mgmt_v3 lock " + "request for operation " + "'Volume %s' to peer %s", + gd_op_list[opinfo.op], + peerinfo->hostname); + dict_unref (dict); + continue; + } + pending_count++; } - pending_count++; } } opinfo.pending_count = pending_count; if (!opinfo.pending_count) - ret = glusterd_op_sm_inject_all_acc (); + ret = glusterd_op_sm_inject_all_acc (&event->txn_id); +out: gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); - return ret; } @@ -2244,17 +2624,12 @@ glusterd_op_ac_send_unlock (glusterd_op_sm_event_t *event, void *ctx) xlator_t *this = NULL; glusterd_peerinfo_t *peerinfo = NULL; uint32_t pending_count = 0; + dict_t *dict = NULL; this = THIS; priv = this->private; GF_ASSERT (priv); - /*ret = glusterd_unlock (MY_UUID); - - if (ret) - goto out; - */ - list_for_each_entry (peerinfo, &priv->peers, uuid_list) { GF_ASSERT (peerinfo); @@ -2264,29 +2639,62 @@ glusterd_op_ac_send_unlock (glusterd_op_sm_event_t *event, void *ctx) (glusterd_op_get_op() != GD_OP_SYNC_VOLUME)) continue; - proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_UNLOCK]; - if (proc->fn) { - ret = proc->fn (NULL, this, peerinfo); - if (ret) { - gf_log (this->name, GF_LOG_WARNING, "Failed to " - "send unlock request for operation " - "'Volume %s' to peer %s", - gd_op_list[opinfo.op], - peerinfo->hostname); - continue; + /* Based on the op_version, release the * + * cluster or mgmt_v3 lock */ + if (priv->op_version < 3) { + proc = &peerinfo->mgmt->proctable[GLUSTERD_MGMT_CLUSTER_UNLOCK]; + if (proc->fn) { + ret = proc->fn (NULL, this, peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Failed to send unlock request " + "for operation 'Volume %s' to " + "peer %s", + gd_op_list[opinfo.op], + peerinfo->hostname); + continue; + } + pending_count++; + } + } else { + dict = glusterd_op_get_ctx (); + dict_ref (dict); + + proc = &peerinfo->mgmt_v3->proctable + [GLUSTERD_MGMT_V3_UNLOCK]; + if (proc->fn) { + ret = dict_set_static_ptr (dict, "peerinfo", + peerinfo); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "failed to set peerinfo"); + dict_unref (dict); + goto out; + } + + ret = proc->fn (NULL, this, dict); + if (ret) { + gf_log (this->name, GF_LOG_WARNING, + "Failed to send mgmt_v3 unlock " + "request for operation " + "'Volume %s' to peer %s", + gd_op_list[opinfo.op], + peerinfo->hostname); + dict_unref (dict); + continue; + } + pending_count++; } - pending_count++; } } opinfo.pending_count = pending_count; if (!opinfo.pending_count) - ret = glusterd_op_sm_inject_all_acc (); + ret = glusterd_op_sm_inject_all_acc (&event->txn_id); +out: gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); - return ret; - } static int @@ -2298,7 +2706,8 @@ glusterd_op_ac_ack_drain (glusterd_op_sm_event_t *event, void *ctx) opinfo.pending_count--; if (!opinfo.pending_count) - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, NULL); gf_log (THIS->name, GF_LOG_DEBUG, "Returning with %d", ret); @@ -2314,43 +2723,94 @@ glusterd_op_ac_send_unlock_drain (glusterd_op_sm_event_t *event, void *ctx) static int glusterd_op_ac_lock (glusterd_op_sm_event_t *event, void *ctx) { - glusterd_op_lock_ctx_t *lock_ctx = NULL; - int32_t ret = 0; + int32_t ret = 0; + char *volname = NULL; + glusterd_op_lock_ctx_t *lock_ctx = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; GF_ASSERT (event); GF_ASSERT (ctx); + this = THIS; + priv = this->private; + lock_ctx = (glusterd_op_lock_ctx_t *)ctx; - ret = glusterd_lock (lock_ctx->uuid); + /* If the req came from a node running on older op_version + * the dict won't be present. Based on it acquiring a cluster + * or mgmt_v3 lock */ + if (lock_ctx->dict == NULL) { + ret = glusterd_lock (lock_ctx->uuid); + glusterd_op_lock_send_resp (lock_ctx->req, ret); + } else { + ret = dict_get_str (lock_ctx->dict, "volname", &volname); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire volname"); + else { + ret = glusterd_mgmt_v3_lock (volname, lock_ctx->uuid, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire lock for %s", + volname); + } - gf_log (THIS->name, GF_LOG_DEBUG, "Lock Returned %d", ret); + glusterd_op_mgmt_v3_lock_send_resp (lock_ctx->req, + &event->txn_id, ret); - glusterd_op_lock_send_resp (lock_ctx->req, ret); + dict_unref (lock_ctx->dict); + } + gf_log (THIS->name, GF_LOG_DEBUG, "Lock Returned %d", ret); return ret; } static int glusterd_op_ac_unlock (glusterd_op_sm_event_t *event, void *ctx) { - int ret = 0; - glusterd_op_lock_ctx_t *lock_ctx = NULL; - xlator_t *this = NULL; - glusterd_conf_t *priv = NULL; + int32_t ret = 0; + char *volname = NULL; + glusterd_op_lock_ctx_t *lock_ctx = NULL; + glusterd_conf_t *priv = NULL; + xlator_t *this = NULL; + GF_ASSERT (event); GF_ASSERT (ctx); this = THIS; priv = this->private; + lock_ctx = (glusterd_op_lock_ctx_t *)ctx; - ret = glusterd_unlock (lock_ctx->uuid); + /* If the req came from a node running on older op_version + * the dict won't be present. Based on it releasing the cluster + * or mgmt_v3 lock */ + if (lock_ctx->dict == NULL) { + ret = glusterd_unlock (lock_ctx->uuid); + glusterd_op_unlock_send_resp (lock_ctx->req, ret); + } else { + ret = dict_get_str (lock_ctx->dict, "volname", &volname); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire volname"); + else { + ret = glusterd_mgmt_v3_unlock (volname, lock_ctx->uuid, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to release lock for %s", volname); + } + + glusterd_op_mgmt_v3_unlock_send_resp (lock_ctx->req, + &event->txn_id, ret); - gf_log (this->name, GF_LOG_DEBUG, "Unlock Returned %d", ret); + dict_unref (lock_ctx->dict); + } - glusterd_op_unlock_send_resp (lock_ctx->req, ret); + gf_log (this->name, GF_LOG_DEBUG, "Unlock Returned %d", ret); if (priv->pending_quorum_action) glusterd_do_quorum_action (); @@ -2388,7 +2848,8 @@ glusterd_op_ac_rcvd_lock_acc (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, + &event->txn_id, NULL); gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -2496,12 +2957,13 @@ glusterd_op_build_payload (dict_t **req, char **op_errstr, dict_t *op_ctx) } break; + case GD_OP_GSYNC_CREATE: case GD_OP_GSYNC_SET: { ret = glusterd_op_gsync_args_get (dict, &errstr, &volname, - NULL); + NULL, NULL); if (ret == 0) { ret = glusterd_dict_set_volid (dict, volname, op_errstr); @@ -2591,9 +3053,6 @@ glusterd_op_build_payload (dict_t **req, char **op_errstr, dict_t *op_ctx) case GD_OP_STATEDUMP_VOLUME: case GD_OP_CLEARLOCKS_VOLUME: case GD_OP_DEFRAG_BRICK_VOLUME: -#ifdef HAVE_BD_XLATOR - case GD_OP_BD_OP: -#endif { ret = dict_get_str (dict, "volname", &volname); if (ret) { @@ -2614,6 +3073,18 @@ glusterd_op_build_payload (dict_t **req, char **op_errstr, dict_t *op_ctx) } break; + case GD_OP_COPY_FILE: + { + dict_copy (dict, req_dict); + break; + } + + case GD_OP_SYS_EXEC: + { + dict_copy (dict, req_dict); + break; + } + default: break; } @@ -2810,7 +3281,8 @@ out: if (dict) dict_unref (dict); if (ret) { - glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL); + glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, + &event->txn_id, NULL); opinfo.op_ret = ret; } @@ -2819,7 +3291,7 @@ out: opinfo.pending_count); if (!opinfo.pending_count) - ret = glusterd_op_sm_inject_all_acc (); + ret = glusterd_op_sm_inject_all_acc (&event->txn_id); gf_log (this->name, GF_LOG_DEBUG, "Returning with %d", ret); @@ -2828,10 +3300,10 @@ out: } static int32_t -glusterd_op_start_rb_timer (dict_t *dict) +glusterd_op_start_rb_timer (dict_t *dict, uuid_t *txn_id) { int32_t op = 0; - struct timeval timeout = {0, }; + struct timespec timeout = {0, }; glusterd_conf_t *priv = NULL; int32_t ret = -1; dict_t *rb_ctx = NULL; @@ -2847,12 +3319,12 @@ glusterd_op_start_rb_timer (dict_t *dict) } if (op != GF_REPLACE_OP_START) { - ret = glusterd_op_sm_inject_all_acc (); + ret = glusterd_op_sm_inject_all_acc (txn_id); goto out; } timeout.tv_sec = 5; - timeout.tv_usec = 0; + timeout.tv_nsec = 0; rb_ctx = dict_copy (dict, rb_ctx); @@ -2862,6 +3334,17 @@ glusterd_op_start_rb_timer (dict_t *dict) ret = -1; goto out; } + + ret = dict_set_bin (rb_ctx, "transaction_id", + txn_id, sizeof(*txn_id)); + if (ret) { + gf_log ("", GF_LOG_ERROR, + "Failed to set transaction id."); + goto out; + } else + gf_log ("", GF_LOG_DEBUG, + "transaction_id = %s", uuid_utoa (*txn_id)); + priv->timer = gf_timer_call_after (THIS->ctx, timeout, glusterd_do_replace_brick, (void *) rb_ctx); @@ -2930,6 +3413,97 @@ out: return ret; } +static int +reassign_defrag_status (dict_t *dict, char *key, gf_defrag_status_t *status) +{ + int ret = 0; + + if (!*status) + return ret; + + switch (*status) { + case GF_DEFRAG_STATUS_STARTED: + *status = GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED; + break; + + case GF_DEFRAG_STATUS_STOPPED: + *status = GF_DEFRAG_STATUS_LAYOUT_FIX_STOPPED; + break; + + case GF_DEFRAG_STATUS_COMPLETE: + *status = GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE; + break; + + case GF_DEFRAG_STATUS_FAILED: + *status = GF_DEFRAG_STATUS_LAYOUT_FIX_FAILED; + break; + default: + break; + } + + ret = dict_set_int32(dict, key, *status); + if (ret) + gf_log (THIS->name, GF_LOG_WARNING, + "failed to reset defrag %s in dict", key); + + return ret; +} + +/* Check and reassign the defrag_status enum got from the rebalance process + * of all peers so that the rebalance-status CLI command can display if a + * full-rebalance or just a fix-layout was carried out. + */ +static int +glusterd_op_check_peer_defrag_status (dict_t *dict, int count) +{ + glusterd_volinfo_t *volinfo = NULL; + gf_defrag_status_t status = GF_DEFRAG_STATUS_NOT_STARTED; + char key[256] = {0,}; + char *volname = NULL; + int ret = -1; + int i = 1; + + ret = dict_get_str (dict, "volname", &volname); + if (ret) { + gf_log (THIS->name, GF_LOG_WARNING, "Unable to get volume name"); + goto out; + } + + ret = glusterd_volinfo_find (volname, &volinfo); + if (ret) { + gf_log (THIS->name, GF_LOG_WARNING, FMTSTR_CHECK_VOL_EXISTS, + volname); + goto out; + } + + if (volinfo->rebal.defrag_cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) { + /* Fix layout was not issued; we don't need to reassign + the status */ + ret = 0; + goto out; + } + + do { + memset (key, 0, 256); + snprintf (key, 256, "status-%d", i); + ret = dict_get_int32 (dict, key, (int32_t *)&status); + if (ret) { + gf_log (THIS->name, GF_LOG_WARNING, + "failed to get defrag %s", key); + goto out; + } + ret = reassign_defrag_status (dict, key, &status); + if (ret) + goto out; + i++; + } while (i <= count); + + ret = 0; +out: + return ret; + +} + /* This function is used to modify the op_ctx dict before sending it back * to cli. This is useful in situations like changing the peer uuids to * hostnames etc. @@ -3033,12 +3607,38 @@ glusterd_op_modify_op_ctx (glusterd_op_t op, void *ctx) goto out; } + /* add 'node-name-%d' into op_ctx with value uuid_str. + this will be used to convert to hostname later */ + { + char key[1024]; + char *uuid_str = NULL; + int i; + + for (i = 1; i <= count; i++) { + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), "node-uuid-%d", i); + ret = dict_get_str (op_ctx, key, &uuid_str); + if (!ret) { + memset (key, 0, sizeof (key)); + snprintf (key, sizeof (key), + "node-name-%d", i); + ret = dict_set_str (op_ctx, key, + uuid_str); + } + } + } + ret = glusterd_op_volume_dict_uuid_to_hostname (op_ctx, - "node-uuid-%d", + "node-name-%d", 1, (count + 1)); if (ret) gf_log (this->name, GF_LOG_WARNING, "Failed uuid to hostname conversion"); + + ret = glusterd_op_check_peer_defrag_status (op_ctx, count); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to reset defrag status for fix-layout"); break; default: @@ -3191,17 +3791,19 @@ out: if (dict) dict_unref (dict); if (ret) { - glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL); + glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, + &event->txn_id, NULL); opinfo.op_ret = ret; } if (!opinfo.pending_count) { if (op == GD_OP_REPLACE_BRICK) { - ret = glusterd_op_start_rb_timer (op_dict); + ret = glusterd_op_start_rb_timer (op_dict, + &event->txn_id); } else { glusterd_op_modify_op_ctx (op, NULL); - ret = glusterd_op_sm_inject_all_acc (); + ret = glusterd_op_sm_inject_all_acc (&event->txn_id); } goto err; } @@ -3226,7 +3828,8 @@ glusterd_op_ac_rcvd_stage_op_acc (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_STAGE_ACC, + &event->txn_id, NULL); out: gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -3247,7 +3850,8 @@ glusterd_op_ac_stage_op_failed (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, NULL); out: gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -3268,7 +3872,8 @@ glusterd_op_ac_commit_op_failed (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, NULL); out: gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -3311,7 +3916,8 @@ glusterd_op_ac_brick_op_failed (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.brick_pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, ev_ctx->commit_ctx); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, ev_ctx->commit_ctx); out: if (ev_ctx->rsp_dict) @@ -3353,7 +3959,7 @@ glusterd_op_ac_rcvd_commit_op_acc (glusterd_op_sm_event_t *event, void *ctx) goto out; } - ret = glusterd_op_start_rb_timer (op_ctx); + ret = glusterd_op_start_rb_timer (op_ctx, &event->txn_id); if (ret) { gf_log (this->name, GF_LOG_ERROR, "Couldn't start " "replace-brick operation."); @@ -3368,10 +3974,12 @@ glusterd_op_ac_rcvd_commit_op_acc (glusterd_op_sm_event_t *event, void *ctx) out: if (commit_ack_inject) { if (ret) - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_RCVD_RJT, + &event->txn_id, NULL); else if (!opinfo.pending_count) { glusterd_op_modify_op_ctx (op, NULL); - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_COMMIT_ACC, + &event->txn_id, NULL); } /*else do nothing*/ } @@ -3392,7 +4000,8 @@ glusterd_op_ac_rcvd_unlock_acc (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, NULL); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACC, + &event->txn_id, NULL); gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -3426,7 +4035,7 @@ glusterd_op_reset_ctx () } int32_t -glusterd_op_txn_complete () +glusterd_op_txn_complete (uuid_t *txn_id) { int32_t ret = -1; glusterd_conf_t *priv = NULL; @@ -3436,6 +4045,7 @@ glusterd_op_txn_complete () rpcsvc_request_t *req = NULL; void *ctx = NULL; char *op_errstr = NULL; + char *volname = NULL; xlator_t *this = NULL; this = THIS; @@ -3458,14 +4068,29 @@ glusterd_op_txn_complete () glusterd_op_reset_ctx (); glusterd_op_clear_errstr (); - ret = glusterd_unlock (MY_UUID); - - /* unlock cant/shouldnt fail here!! */ - if (ret) { - gf_log (this->name, GF_LOG_CRITICAL, - "Unable to clear local lock, ret: %d", ret); + /* Based on the op-version, we release the cluster or mgmt_v3 lock */ + if (priv->op_version < 3) { + ret = glusterd_unlock (MY_UUID); + /* unlock cant/shouldnt fail here!! */ + if (ret) + gf_log (this->name, GF_LOG_CRITICAL, + "Unable to clear local lock, ret: %d", ret); + else + gf_log (this->name, GF_LOG_DEBUG, "Cleared local lock"); } else { - gf_log (this->name, GF_LOG_DEBUG, "Cleared local lock"); + ret = dict_get_str (ctx, "volname", &volname); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to acquire volname"); + + if (volname) { + ret = glusterd_mgmt_v3_unlock (volname, MY_UUID, + "vol"); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to release lock for %s", + volname); + } } ret = glusterd_op_send_cli_response (op, op_ret, @@ -3484,6 +4109,13 @@ glusterd_op_txn_complete () if (priv->pending_quorum_action) glusterd_do_quorum_action (); + + /* Clearing the transaction opinfo */ + ret = glusterd_clear_txn_opinfo (txn_id); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to clear transaction's opinfo"); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } @@ -3495,7 +4127,7 @@ glusterd_op_ac_unlocked_all (glusterd_op_sm_event_t *event, void *ctx) GF_ASSERT (event); - ret = glusterd_op_txn_complete (); + ret = glusterd_op_txn_complete (&event->txn_id); gf_log (THIS->name, GF_LOG_DEBUG, "Returning %d", ret); @@ -3512,6 +4144,7 @@ glusterd_op_ac_stage_op (glusterd_op_sm_event_t *event, void *ctx) char *op_errstr = NULL; dict_t *dict = NULL; xlator_t *this = NULL; + uuid_t *txn_id = NULL; this = THIS; GF_ASSERT (this); @@ -3537,6 +4170,19 @@ glusterd_op_ac_stage_op (glusterd_op_sm_event_t *event, void *ctx) status); } + txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + + if (txn_id) + uuid_copy (*txn_id, event->txn_id); + else + gf_log (this->name, GF_LOG_ERROR, "Out of Memory"); + + ret = dict_set_bin (rsp_dict, "transaction_id", + txn_id, sizeof(*txn_id)); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to set transaction id."); + ret = glusterd_op_stage_send_resp (req_ctx->req, req_ctx->op, status, op_errstr, rsp_dict); @@ -3601,6 +4247,7 @@ glusterd_op_ac_commit_op (glusterd_op_sm_event_t *event, void *ctx) dict_t *dict = NULL; dict_t *rsp_dict = NULL; xlator_t *this = NULL; + uuid_t *txn_id = NULL; this = THIS; GF_ASSERT (this); @@ -3630,10 +4277,22 @@ glusterd_op_ac_commit_op (glusterd_op_sm_event_t *event, void *ctx) "'Volume %s' failed: %d", gd_op_list[req_ctx->op], status); + txn_id = GF_CALLOC (1, sizeof(uuid_t), gf_common_mt_uuid_t); + + if (txn_id) + uuid_copy (*txn_id, event->txn_id); + else + gf_log (this->name, GF_LOG_ERROR, "Out of Memory"); + + ret = dict_set_bin (rsp_dict, "transaction_id", + txn_id, sizeof(*txn_id)); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Failed to set transaction id."); + ret = glusterd_op_commit_send_resp (req_ctx->req, req_ctx->op, status, op_errstr, rsp_dict); - glusterd_op_fini_ctx (); if (op_errstr && (strcmp (op_errstr, ""))) GF_FREE (op_errstr); @@ -3662,7 +4321,6 @@ glusterd_op_ac_send_commit_failed (glusterd_op_sm_event_t *event, void *ctx) opinfo.op_ret, opinfo.op_errstr, op_ctx); - glusterd_op_fini_ctx (); if (opinfo.op_errstr && (strcmp (opinfo.op_errstr, ""))) { GF_FREE (opinfo.op_errstr); opinfo.op_errstr = NULL; @@ -3747,6 +4405,10 @@ glusterd_op_stage_validate (glusterd_op_t op, dict_t *dict, char **op_errstr, ret = glusterd_op_stage_sync_volume (dict, op_errstr); break; + case GD_OP_GSYNC_CREATE: + ret = glusterd_op_stage_gsync_create (dict, op_errstr); + break; + case GD_OP_GSYNC_SET: ret = glusterd_op_stage_gsync_set (dict, op_errstr); break; @@ -3780,18 +4442,21 @@ glusterd_op_stage_validate (glusterd_op_t op, dict_t *dict, char **op_errstr, ret = glusterd_op_stage_clearlocks_volume (dict, op_errstr); break; -#ifdef HAVE_BD_XLATOR - case GD_OP_BD_OP: - ret = glusterd_op_stage_bd (dict, op_errstr); + + case GD_OP_COPY_FILE: + ret = glusterd_op_stage_copy_file (dict, op_errstr); break; -#endif + + case GD_OP_SYS_EXEC: + ret = glusterd_op_stage_sys_exec (dict, op_errstr); + break; + default: gf_log (this->name, GF_LOG_ERROR, "Unknown op %s", gd_op_list[op]); } - gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); - + gf_log (this->name, GF_LOG_DEBUG, "OP = %d. Returning %d", op, ret); return ret; } @@ -3849,6 +4514,11 @@ glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr, ret = glusterd_op_sync_volume (dict, op_errstr, rsp_dict); break; + case GD_OP_GSYNC_CREATE: + ret = glusterd_op_gsync_create (dict, op_errstr, + rsp_dict); + break; + case GD_OP_GSYNC_SET: ret = glusterd_op_gsync_set (dict, op_errstr, rsp_dict); break; @@ -3883,11 +4553,15 @@ glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr, ret = glusterd_op_clearlocks_volume (dict, op_errstr, rsp_dict); break; -#ifdef HAVE_BD_XLATOR - case GD_OP_BD_OP: - ret = 0; + + case GD_OP_COPY_FILE: + ret = glusterd_op_copy_file (dict, op_errstr); break; -#endif + + case GD_OP_SYS_EXEC: + ret = glusterd_op_sys_exec (dict, op_errstr, rsp_dict); + break; + default: gf_log (this->name, GF_LOG_ERROR, "Unknown op %s", gd_op_list[op]); @@ -3896,11 +4570,12 @@ glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr, if (ret == 0) glusterd_op_commit_hook (op, dict, GD_COMMIT_HOOK_POST); - gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); + gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); return ret; } + static int glusterd_bricks_select_stop_volume (dict_t *dict, char **op_errstr, struct list_head *selected) @@ -4217,24 +4892,95 @@ out: } int +get_replica_index_for_per_replica_cmd (glusterd_volinfo_t *volinfo, + dict_t *dict) { + int ret = 0; + char *hostname = NULL; + char *path = NULL; + int index = 0; + glusterd_brickinfo_t *brickinfo = NULL; + int cmd_replica_index = -1; + int replica_count = -1; + + + if (!dict) { + ret = -1; + goto out; + } + + ret = dict_get_str (dict, "per-replica-cmd-hostname", &hostname); + if (ret) + goto out; + ret = dict_get_str (dict, "per-replica-cmd-path", &path); + if (ret) + goto out; + + replica_count = volinfo->replica_count; + + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { + if (uuid_is_null (brickinfo->uuid)) + (void)glusterd_resolve_brick (brickinfo); + if (!strcmp (brickinfo->path, path) && + !strcmp (brickinfo->hostname, hostname)) { + cmd_replica_index = index/(replica_count); + goto out; + } + index++; + } + + +out: + if (ret) + cmd_replica_index = -1; + + return cmd_replica_index; +} + +int _select_rxlators_with_local_bricks (xlator_t *this, glusterd_volinfo_t *volinfo, - dict_t *dict) + dict_t *dict, cli_cmd_type type) { glusterd_brickinfo_t *brickinfo = NULL; glusterd_conf_t *priv = NULL; - int index = 1; + int index = 0; int rxlator_count = 0; int replica_count = 0; gf_boolean_t add = _gf_false; + int ret = 0; + int cmd_replica_index = -1; priv = this->private; replica_count = volinfo->replica_count; + + if (type == PER_REPLICA) { + + cmd_replica_index = get_replica_index_for_per_replica_cmd + (volinfo, dict); + if (cmd_replica_index == -1) { + ret = -1; + goto err; + } + } + + index = 1; + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { if (uuid_is_null (brickinfo->uuid)) (void)glusterd_resolve_brick (brickinfo); - if (!uuid_compare (MY_UUID, brickinfo->uuid)) - add = _gf_true; + switch (type) { + case ALL_REPLICA: + if (!uuid_compare (MY_UUID, brickinfo->uuid)) + add = _gf_true; + break; + case PER_REPLICA: + if (!uuid_compare (MY_UUID, brickinfo->uuid) && + ((index-1)/replica_count == cmd_replica_index)) + + add = _gf_true; + break; + } + if (index % replica_count == 0) { if (add) { _add_rxlator_to_dict (dict, volinfo->volname, @@ -4247,6 +4993,10 @@ _select_rxlators_with_local_bricks (xlator_t *this, glusterd_volinfo_t *volinfo, index++; } +err: + if (ret) + rxlator_count = -1; + return rxlator_count; } @@ -4287,9 +5037,10 @@ _select_rxlators_for_full_self_heal (xlator_t *this, return rxlator_count; } -#ifdef HAVE_BD_XLATOR + static int -glusterd_bricks_select_bd (dict_t *dict, char **op_errstr) +glusterd_bricks_select_snap (dict_t *dict, char **op_errstr, + struct list_head *selected) { int ret = -1; glusterd_conf_t *priv = NULL; @@ -4307,31 +5058,31 @@ glusterd_bricks_select_bd (dict_t *dict, char **op_errstr) ret = dict_get_str (dict, "volname", &volname); if (ret) { - gf_log (this->name, GF_LOG_ERROR, "Unable to get volname"); + gf_log (this->name, GF_LOG_ERROR, "Unable to get" + " volname"); goto out; } ret = glusterd_volinfo_find (volname, &volinfo); if (ret) goto out; - pending_node = GF_CALLOC (1, sizeof (*pending_node), - gf_gld_mt_pending_node_t); - if (!pending_node) { - ret = -1; - goto out; - } - list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { brick_index++; if (uuid_compare (brickinfo->uuid, MY_UUID) || !glusterd_is_brick_started (brickinfo)) { continue; } + pending_node = GF_CALLOC (1, sizeof (*pending_node), + gf_gld_mt_pending_node_t); + if (!pending_node) { + ret = -1; + goto out; + } pending_node->node = brickinfo; pending_node->type = GD_NODE_BRICK; pending_node->index = brick_index; list_add_tail (&pending_node->list, - &opinfo.pending_bricks); + selected); pending_node = NULL; } @@ -4341,10 +5092,10 @@ out: gf_log (THIS->name, GF_LOG_DEBUG, "Returning ret %d", ret); return ret; } -#endif static int -fill_shd_status_for_local_bricks (dict_t *dict, glusterd_volinfo_t *volinfo) +fill_shd_status_for_local_bricks (dict_t *dict, glusterd_volinfo_t *volinfo, + cli_cmd_type type, dict_t *req_dict) { glusterd_brickinfo_t *brickinfo = NULL; char msg[1024] = {0,}; @@ -4353,10 +5104,22 @@ fill_shd_status_for_local_bricks (dict_t *dict, glusterd_volinfo_t *volinfo) int index = 0; int ret = 0; xlator_t *this = NULL; + int cmd_replica_index = -1; this = THIS; snprintf (msg, sizeof (msg), "self-heal-daemon is not running on"); + if (type == PER_REPLICA) { + cmd_replica_index = get_replica_index_for_per_replica_cmd + (volinfo, req_dict); + if (cmd_replica_index == -1) { + gf_log (THIS->name, GF_LOG_ERROR, "Could not find the " + "replica index for per replica type command"); + ret = -1; + goto out; + } + } + list_for_each_entry (brickinfo, &volinfo->bricks, brick_list) { if (uuid_is_null (brickinfo->uuid)) (void)glusterd_resolve_brick (brickinfo); @@ -4365,6 +5128,14 @@ fill_shd_status_for_local_bricks (dict_t *dict, glusterd_volinfo_t *volinfo) index++; continue; } + + if (type == PER_REPLICA) { + if (cmd_replica_index != (index/volinfo->replica_count)) { + index++; + continue; + } + + } snprintf (key, sizeof (key), "%d-status",index); snprintf (value, sizeof (value), "%s %s",msg, uuid_utoa(MY_UUID)); @@ -4433,21 +5204,49 @@ glusterd_bricks_select_heal_volume (dict_t *dict, char **op_errstr, goto out; } + switch (heal_op) { + case GF_AFR_OP_INDEX_SUMMARY: + case GF_AFR_OP_STATISTICS_HEAL_COUNT: + if (!glusterd_is_nodesvc_online ("glustershd")) { + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, "Received " + "empty ctx."); + goto out; + } - if (!glusterd_is_nodesvc_online ("glustershd") && - (heal_op == GF_AFR_OP_INDEX_SUMMARY)) { - - if (!rsp_dict) { - gf_log (this->name, GF_LOG_ERROR, "Received empty " - "ctx."); + ret = fill_shd_status_for_local_bricks (rsp_dict, + volinfo, + ALL_REPLICA, + dict); + if (ret) + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "fill the shd status for the local " + "bricks"); goto out; + } + break; + case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA: + if (!glusterd_is_nodesvc_online ("glustershd")) { + if (!rsp_dict) { + gf_log (this->name, GF_LOG_ERROR, "Received " + "empty ctx."); + goto out; + } + ret = fill_shd_status_for_local_bricks (rsp_dict, + volinfo, + PER_REPLICA, + dict); + if (ret) + gf_log (this->name, GF_LOG_ERROR, "Unable to " + "fill the shd status for the local" + " bricks."); + goto out; - ret = fill_shd_status_for_local_bricks (rsp_dict, volinfo); - if (ret) - gf_log (this->name, GF_LOG_ERROR, "Unable to fill the shd" - " status for the local bricks"); - goto out; + } + break; + default: + break; } @@ -4457,14 +5256,28 @@ glusterd_bricks_select_heal_volume (dict_t *dict, char **op_errstr, volinfo, dict); break; + case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA: + rxlator_count = _select_rxlators_with_local_bricks (this, + volinfo, + dict, + PER_REPLICA); + break; default: rxlator_count = _select_rxlators_with_local_bricks (this, volinfo, - dict); + dict, + ALL_REPLICA); break; } if (!rxlator_count) goto out; + if (rxlator_count == -1){ + gf_log (this->name, GF_LOG_ERROR, "Could not determine the" + "translator count"); + ret = -1; + goto out; + } + ret = dict_set_int32 (dict, "count", rxlator_count); if (ret) goto out; @@ -4726,7 +5539,8 @@ glusterd_op_ac_send_brick_op (glusterd_op_sm_event_t *event, void *ctx) if (!opinfo.pending_count && !opinfo.brick_pending_count) { glusterd_clear_pending_nodes (&opinfo.pending_bricks); - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, req_ctx); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, + &event->txn_id, req_ctx); } out: @@ -4780,7 +5594,8 @@ glusterd_op_ac_rcvd_brick_op_acc (glusterd_op_sm_event_t *event, void *ctx) if (opinfo.brick_pending_count > 0) goto out; - ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, ev_ctx->commit_ctx); + ret = glusterd_op_sm_inject_event (GD_OP_EVENT_ALL_ACK, &event->txn_id, + ev_ctx->commit_ctx); out: if (ev_ctx->rsp_dict) @@ -4832,11 +5647,9 @@ glusterd_op_bricks_select (glusterd_op_t op, dict_t *dict, char **op_errstr, ret = glusterd_bricks_select_rebalance_volume (dict, op_errstr, selected); break; -#ifdef HAVE_BD_XLATOR - case GD_OP_BD_OP: - ret = glusterd_bricks_select_bd (dict, op_errstr); + case GD_OP_SNAP: + ret = glusterd_bricks_select_snap (dict, op_errstr, selected); break; -#endif default: break; } @@ -5158,7 +5971,7 @@ glusterd_op_sm_new_event (glusterd_op_sm_event_type_t event_type, int glusterd_op_sm_inject_event (glusterd_op_sm_event_type_t event_type, - void *ctx) + uuid_t *txn_id, void *ctx) { int32_t ret = -1; glusterd_op_sm_event_t *event = NULL; @@ -5173,6 +5986,9 @@ glusterd_op_sm_inject_event (glusterd_op_sm_event_type_t event_type, event->ctx = ctx; + if (txn_id) + uuid_copy (event->txn_id, *txn_id); + gf_log (THIS->name, GF_LOG_DEBUG, "Enqueue event: '%s'", glusterd_op_sm_event_name_get (event->event)); list_add_tail (&event->list, &gd_op_sm_queue); @@ -5233,6 +6049,7 @@ glusterd_op_sm () glusterd_op_sm_t *state = NULL; glusterd_op_sm_event_type_t event_type = GD_OP_EVENT_NONE; xlator_t *this = NULL; + glusterd_op_info_t txn_op_info; this = THIS; GF_ASSERT (this); @@ -5253,6 +6070,20 @@ glusterd_op_sm () "type: '%s'", glusterd_op_sm_event_name_get(event_type)); + gf_log ("", GF_LOG_DEBUG, "transaction ID = %s", + uuid_utoa (event->txn_id)); + + ret = glusterd_get_txn_opinfo (&event->txn_id, + &txn_op_info); + if (ret) { + gf_log (this->name, GF_LOG_ERROR, + "Unable to get transaction's opinfo"); + glusterd_destroy_op_event_ctx (event); + GF_FREE (event); + continue; + } else + opinfo = txn_op_info; + state = glusterd_op_state_table[opinfo.state.state]; GF_ASSERT (state); @@ -5283,8 +6114,27 @@ glusterd_op_sm () return ret; } + if ((state[event_type].next_state == + GD_OP_STATE_DEFAULT) && + (event_type == GD_OP_EVENT_UNLOCK)) { + /* Clearing the transaction opinfo */ + ret = glusterd_clear_txn_opinfo(&event->txn_id); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to clear " + "transaction's opinfo"); + } else { + ret = glusterd_set_txn_opinfo (&event->txn_id, + &opinfo); + if (ret) + gf_log (this->name, GF_LOG_ERROR, + "Unable to set " + "transaction's opinfo"); + } + glusterd_destroy_op_event_ctx (event); GF_FREE (event); + } } @@ -5338,52 +6188,6 @@ glusterd_op_clear_op (glusterd_op_t op) } int32_t -glusterd_op_init_ctx (glusterd_op_t op) -{ - int ret = 0; - dict_t *dict = NULL; - xlator_t *this = NULL; - - this = THIS; - GF_ASSERT (this); - GF_ASSERT (GD_OP_NONE < op && op < GD_OP_MAX); - - if (_gf_false == glusterd_need_brick_op (op)) { - gf_log (this->name, GF_LOG_DEBUG, "Received op: %s, returning", - gd_op_list[op]); - goto out; - } - dict = dict_new (); - if (dict == NULL) { - ret = -1; - goto out; - } - ret = glusterd_op_set_ctx (dict); - if (ret) - goto out; -out: - gf_log (this->name, GF_LOG_DEBUG, "Returning %d", ret); - return ret; -} - - - -int32_t -glusterd_op_fini_ctx () -{ - dict_t *dict = NULL; - - dict = glusterd_op_get_ctx (); - if (dict) - dict_unref (dict); - - glusterd_op_reset_ctx (); - return 0; -} - - - -int32_t glusterd_op_free_ctx (glusterd_op_t op, void *ctx) { @@ -5409,9 +6213,6 @@ glusterd_op_free_ctx (glusterd_op_t op, void *ctx) case GD_OP_STATEDUMP_VOLUME: case GD_OP_CLEARLOCKS_VOLUME: case GD_OP_DEFRAG_BRICK_VOLUME: -#ifdef HAVE_BD_XLATOR - case GD_OP_BD_OP: -#endif dict_unref (ctx); break; default: @@ -5440,4 +6241,3 @@ glusterd_op_sm_init () pthread_mutex_init (&gd_op_sm_lock, NULL); return 0; } - |