From 25daa42911d2ff697880ee29c591cac5f2abebed Mon Sep 17 00:00:00 2001 From: Amar Tumballi Date: Fri, 9 Sep 2011 09:42:51 +0530 Subject: support for de-commissioning a node using 'remove-brick' to achieve this, we now create volume-file with 'decommissioned-nodes' option in distribute volume, then just perform the rebalance set of operations (with 'force' flag set). now onwards, the 'remove-brick' (with 'start' option) operation tries to migrate data from removed bricks to existing bricks. 'remove-brick' also supports similar options as of replace-brick. * (no options) -> works as 'force', will have the current behavior of remove-brick, ie., no data-migration, volume changes. * start (starts remove-brick with data-migration/draining process, which takes care of migrating data and once complete, will commit the changes to volume file) * pause (stop data migration, but keep the volume file intact with extra options whatever is set) * abort (stop data-migration, and fall back to old configuration) * commit (if volume is stopped, commits the changes to volumefile) * force (stops the data-migration and commits the changes to volume file) Change-Id: I3952bcfbe604a0952e68b6accace7014d5e401d3 BUG: 1952 Reviewed-on: http://review.gluster.com/118 Tested-by: Gluster Build System Reviewed-by: Vijay Bellur --- xlators/mgmt/glusterd/src/glusterd-brick-ops.c | 369 +++++++++++++++++---- xlators/mgmt/glusterd/src/glusterd-op-sm.c | 11 +- xlators/mgmt/glusterd/src/glusterd-rebalance.c | 46 ++- xlators/mgmt/glusterd/src/glusterd-replace-brick.c | 2 +- xlators/mgmt/glusterd/src/glusterd-store.c | 7 + xlators/mgmt/glusterd/src/glusterd-store.h | 1 + xlators/mgmt/glusterd/src/glusterd-utils.c | 2 + xlators/mgmt/glusterd/src/glusterd-volgen.c | 104 ++++-- xlators/mgmt/glusterd/src/glusterd.h | 34 +- 9 files changed, 446 insertions(+), 130 deletions(-) (limited to 'xlators/mgmt') diff --git a/xlators/mgmt/glusterd/src/glusterd-brick-ops.c b/xlators/mgmt/glusterd/src/glusterd-brick-ops.c index 8b3a03b6f..8832c69ed 100644 --- a/xlators/mgmt/glusterd/src/glusterd-brick-ops.c +++ b/xlators/mgmt/glusterd/src/glusterd-brick-ops.c @@ -312,7 +312,8 @@ glusterd_handle_remove_brick (rpcsvc_request_t *req) strcpy (vol_type, "distribute"); /* Do not allow remove-brick if the volume is plain stripe */ - if ((volinfo->type == GF_CLUSTER_TYPE_STRIPE) && (volinfo->brick_count == volinfo->sub_count)) { + if ((volinfo->type == GF_CLUSTER_TYPE_STRIPE) && + (volinfo->brick_count == volinfo->sub_count)) { snprintf (err_str, 2048, "Removing brick from a plain stripe is not allowed"); gf_log ("glusterd", GF_LOG_ERROR, "%s", err_str); ret = -1; @@ -321,8 +322,8 @@ glusterd_handle_remove_brick (rpcsvc_request_t *req) /* Do not allow remove-brick if the bricks given is less than the replica count or stripe count */ - if (((volinfo->type == GF_CLUSTER_TYPE_REPLICATE) || (volinfo->type == GF_CLUSTER_TYPE_STRIPE)) - && !(volinfo->brick_count <= volinfo->sub_count)) { + if ((volinfo->type != GF_CLUSTER_TYPE_NONE) && + !(volinfo->brick_count <= volinfo->sub_count)) { if (volinfo->sub_count && (count % volinfo->sub_count != 0)) { snprintf (err_str, 2048, "Remove brick incorrect" " brick count of %d for %s %d", @@ -512,16 +513,20 @@ out: int -glusterd_op_perform_remove_brick (glusterd_volinfo_t *volinfo, char *brick) +glusterd_op_perform_remove_brick (glusterd_volinfo_t *volinfo, char *brick, + int force, int *need_migrate) { - glusterd_brickinfo_t *brickinfo = NULL; char *dup_brick = NULL; - int32_t ret = -1; + int32_t ret = -1; + glusterd_conf_t *priv = NULL; GF_ASSERT (volinfo); GF_ASSERT (brick); + priv = THIS->private; + GF_ASSERT (priv); + dup_brick = gf_strdup (brick); if (!dup_brick) goto out; @@ -534,15 +539,26 @@ glusterd_op_perform_remove_brick (glusterd_volinfo_t *volinfo, char *brick) if (ret) goto out; - if (GLUSTERD_STATUS_STARTED == volinfo->status) { - ret = glusterd_brick_stop (volinfo, brickinfo); - if (ret) { - gf_log ("", GF_LOG_ERROR, "Unable to stop " - "glusterfs, ret: %d", ret); - goto out; + if (!uuid_compare (brickinfo->uuid, priv->uuid)) { + /* Only if the brick is in this glusterd, do the rebalance */ + if (need_migrate) + *need_migrate = 1; + } + + if (force) { + if (GLUSTERD_STATUS_STARTED == volinfo->status) { + ret = glusterd_brick_stop (volinfo, brickinfo); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, "Unable to stop " + "glusterfs, ret: %d", ret); + goto out; + } } + glusterd_delete_brick (volinfo, brickinfo); + goto out; } - glusterd_delete_brick (volinfo, brickinfo); + + brickinfo->decommissioned = 1; out: if (dup_brick) GF_FREE (dup_brick); @@ -700,17 +716,18 @@ out: } int -glusterd_op_stage_remove_brick (dict_t *dict) +glusterd_op_stage_remove_brick (dict_t *dict, char **op_errstr) { - int ret = -1; - char *volname = NULL; - glusterd_volinfo_t *volinfo = NULL; - dict_t *ctx = NULL; - char *errstr = NULL; - int32_t brick_count = 0; + int ret = -1; + char *volname = NULL; + glusterd_volinfo_t *volinfo = NULL; + char *errstr = NULL; + int32_t brick_count = 0; + char msg[2048] = {0,}; + int32_t flag = 0; + gf1_op_commands cmd = GF_OP_CMD_NONE; ret = dict_get_str (dict, "volname", &volname); - if (ret) { gf_log ("", GF_LOG_ERROR, "Unable to get volume name"); goto out; @@ -723,25 +740,64 @@ glusterd_op_stage_remove_brick (dict_t *dict) goto out; } - if (glusterd_is_defrag_on(volinfo)) { - ctx = glusterd_op_get_ctx (); - errstr = gf_strdup("Rebalance is in progress. Please retry" - " after completion"); - if (!errstr) { - ret = -1; + ret = dict_get_int32 (dict, "command", &flag); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to get brick count"); + goto out; + } + cmd = flag; + + ret = -1; + switch (cmd) { + case GF_OP_CMD_NONE: + errstr = gf_strdup ("no remove-brick command issued"); + goto out; + + case GF_OP_CMD_STATUS: + ret = 0; + goto out; + + case GF_OP_CMD_START: + { + if (GLUSTERD_STATUS_STARTED != volinfo->status) { + snprintf (msg, sizeof (msg), "Volume %s needs to be started " + "before remove-brick (you can use 'force' or " + "'commit' to override this behavior)", + volinfo->volname); + errstr = gf_strdup (msg); + gf_log (THIS->name, GF_LOG_ERROR, "%s", errstr); goto out; } - gf_log ("glusterd", GF_LOG_ERROR, "%s", errstr); - ret = dict_set_dynstr (ctx, "errstr", errstr); - if (ret) { - GF_FREE (errstr); - gf_log ("", GF_LOG_DEBUG, - "failed to set errstr ctx"); + if (glusterd_is_defrag_on(volinfo)) { + errstr = gf_strdup("Rebalance is in progress. Please retry" + " after completion"); + gf_log ("glusterd", GF_LOG_ERROR, "%s", errstr); goto out; } + break; + } - ret = -1; - goto out; + case GF_OP_CMD_PAUSE: + case GF_OP_CMD_ABORT: + { + if (!volinfo->decommission_in_progress) { + errstr = gf_strdup("remove-brick is not in progress"); + gf_log ("glusterd", GF_LOG_ERROR, "%s", errstr); + goto out; + } + break; + } + + case GF_OP_CMD_COMMIT: + if (volinfo->decommission_in_progress) { + errstr = gf_strdup ("use 'force' option as migration " + "is in progress"); + goto out; + } + break; + + case GF_OP_CMD_COMMIT_FORCE: + break; } ret = dict_get_int32 (dict, "count", &brick_count); @@ -750,40 +806,95 @@ glusterd_op_stage_remove_brick (dict_t *dict) goto out; } + ret = 0; if (volinfo->brick_count == brick_count) { - ctx = glusterd_op_get_ctx (); - if (!ctx) { - gf_log ("", GF_LOG_ERROR, - "Operation Context is not present"); - ret = -1; - goto out; - } errstr = gf_strdup ("Deleting all the bricks of the " "volume is not allowed"); - if (!errstr) { - gf_log ("", GF_LOG_ERROR, "Out of memory"); - ret = -1; - goto out; - } - - ret = dict_set_dynstr (ctx, "errstr", errstr); - if (ret) { - GF_FREE (errstr); - gf_log ("", GF_LOG_DEBUG, - "failed to set pump status in ctx"); - goto out; - } - ret = -1; goto out; } out: gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); + if (ret && errstr) { + if (op_errstr) + *op_errstr = errstr; + } return ret; } +int +glusterd_remove_brick_migrate_cbk (glusterd_volinfo_t *volinfo, + gf_defrag_status_t status) +{ + int ret = 0; + glusterd_brickinfo_t *brickinfo = NULL; + glusterd_brickinfo_t *tmp = NULL; + + switch (status) { + case GF_DEFRAG_STATUS_PAUSED: + case GF_DEFRAG_STATUS_FAILED: + /* No changes required in the volume file. + everything should remain as is */ + break; + case GF_DEFRAG_STATUS_STOPPED: + /* Fall back to the old volume file */ + list_for_each_entry_safe (brickinfo, tmp, &volinfo->bricks, brick_list) { + if (!brickinfo->decommissioned) + continue; + brickinfo->decommissioned = 0; + } + break; + + case GF_DEFRAG_STATUS_COMPLETE: + /* Done with the task, you can remove the brick from the + volume file */ + list_for_each_entry_safe (brickinfo, tmp, &volinfo->bricks, brick_list) { + if (!brickinfo->decommissioned) + continue; + gf_log (THIS->name, GF_LOG_INFO, "removing the brick %s", + brickinfo->path); + brickinfo->decommissioned = 0; + if (GLUSTERD_STATUS_STARTED == volinfo->status) { + ret = glusterd_brick_stop (volinfo, brickinfo); + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "Unable to stop glusterfs (%d)", ret); + } + } + glusterd_delete_brick (volinfo, brickinfo); + } + break; + + default: + GF_ASSERT (!"cbk function called with wrong status"); + break; + } + + ret = glusterd_create_volfiles_and_notify_services (volinfo); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, + "Unable to write volume files (%d)", ret); + + ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, + "Unable to store volume info (%d)", ret); + + + if (GLUSTERD_STATUS_STARTED == volinfo->status) { + ret = glusterd_check_generate_start_nfs (); + if (ret) + gf_log (THIS->name, GF_LOG_ERROR, + "Unable to start nfs process (%d)", ret); + } + + volinfo->decommission_in_progress = 0; + return 0; +} + + int glusterd_op_add_brick (dict_t *dict, char **op_errstr) { @@ -848,15 +959,20 @@ out: } int -glusterd_op_remove_brick (dict_t *dict) +glusterd_op_remove_brick (dict_t *dict, char **op_errstr) { - int ret = -1; - char *volname = NULL; - glusterd_volinfo_t *volinfo = NULL; - char *brick = NULL; - int32_t count = 0; - int32_t i = 1; - char key[256] = {0,}; + int ret = -1; + char *volname = NULL; + glusterd_volinfo_t *volinfo = NULL; + char *brick = NULL; + int32_t count = 0; + int32_t i = 1; + char key[256] = {0,}; + int32_t flag = 0; + char err_str[4096] = {0,}; + int need_rebalance = 0; + int force = 0; + gf1_op_commands cmd = 0; ret = dict_get_str (dict, "volname", &volname); @@ -866,12 +982,99 @@ glusterd_op_remove_brick (dict_t *dict) } ret = glusterd_volinfo_find (volname, &volinfo); - if (ret) { gf_log ("", GF_LOG_ERROR, "Unable to allocate memory"); goto out; } + ret = dict_get_int32 (dict, "command", &flag); + if (ret) { + gf_log ("", GF_LOG_ERROR, "Unable to get brick count"); + goto out; + } + cmd = flag; + + ret = -1; + switch (cmd) { + case GF_OP_CMD_NONE: + goto out; + + case GF_OP_CMD_STATUS: + ret = 0; + goto out; + + case GF_OP_CMD_PAUSE: + { + if (volinfo->decommission_in_progress) { + if (volinfo->defrag == (void *)1) + volinfo->defrag = NULL; + + if (volinfo->defrag) { + LOCK (&volinfo->defrag->lock); + + volinfo->defrag_status = GF_DEFRAG_STATUS_PAUSED; + + UNLOCK (&volinfo->defrag->lock); + } + } + + /* rebalance '_cbk()' will take care of volume file updates */ + ret = 0; + goto out; + } + + case GF_OP_CMD_ABORT: + { + if (volinfo->decommission_in_progress) { + if (volinfo->defrag == (void *)1) + volinfo->defrag = NULL; + + if (volinfo->defrag) { + LOCK (&volinfo->defrag->lock); + + volinfo->defrag_status = GF_DEFRAG_STATUS_STOPPED; + + UNLOCK (&volinfo->defrag->lock); + } + } + + /* rebalance '_cbk()' will take care of volume file updates */ + ret = 0; + goto out; + } + + case GF_OP_CMD_START: + force = 0; + break; + + case GF_OP_CMD_COMMIT: + force = 1; + break; + + case GF_OP_CMD_COMMIT_FORCE: + + if (volinfo->decommission_in_progress) { + if (volinfo->defrag == (void *)1) + volinfo->defrag = NULL; + + if (volinfo->defrag) { + LOCK (&volinfo->defrag->lock); + /* Fake 'rebalance-complete' so the graph change + happens right away */ + volinfo->defrag_status = GF_DEFRAG_STATUS_COMPLETE; + + UNLOCK (&volinfo->defrag->lock); + } + ret = 0; + /* Graph change happens in rebalance _cbk function, + no need to do anything here */ + goto out; + } + + force = 1; + break; + } + ret = dict_get_int32 (dict, "count", &count); if (ret) { gf_log ("", GF_LOG_ERROR, "Unable to get count"); @@ -887,26 +1090,46 @@ glusterd_op_remove_brick (dict_t *dict) goto out; } - ret = glusterd_op_perform_remove_brick (volinfo, brick); + ret = glusterd_op_perform_remove_brick (volinfo, brick, force, + (i == 1) ? &need_rebalance : NULL); if (ret) goto out; i++; } ret = glusterd_create_volfiles_and_notify_services (volinfo); - if (ret) + if (ret) { + gf_log (THIS->name, GF_LOG_WARNING, "failed to create volfiles"); goto out; - - volinfo->defrag_status = 0; + } ret = glusterd_store_volinfo (volinfo, GLUSTERD_VOLINFO_VER_AC_INCREMENT); - - if (ret) + if (ret) { + gf_log (THIS->name, GF_LOG_WARNING, "failed to store volinfo"); goto out; + } - if (GLUSTERD_STATUS_STARTED == volinfo->status) - ret = glusterd_check_generate_start_nfs (); + volinfo->defrag_status = 0; + if (!force && need_rebalance) { + /* perform the rebalance operations */ + ret = glusterd_handle_defrag_start (volinfo, err_str, 4096, + GF_DEFRAG_CMD_START_FORCE, + glusterd_remove_brick_migrate_cbk); + if (!ret) + volinfo->decommission_in_progress = 1; + + if (ret) { + gf_log (THIS->name, GF_LOG_ERROR, + "failed to start the rebalance"); + } + } else { + if (GLUSTERD_STATUS_STARTED == volinfo->status) + ret = glusterd_check_generate_start_nfs (); + } out: + if (ret && err_str[0] && op_errstr) + *op_errstr = gf_strdup (err_str); + return ret; } diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.c b/xlators/mgmt/glusterd/src/glusterd-op-sm.c index 3b30fb080..53556984a 100644 --- a/xlators/mgmt/glusterd/src/glusterd-op-sm.c +++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.c @@ -2297,7 +2297,7 @@ glusterd_op_stage_validate (glusterd_op_t op, dict_t *dict, char **op_errstr, break; case GD_OP_REMOVE_BRICK: - ret = glusterd_op_stage_remove_brick (dict); + ret = glusterd_op_stage_remove_brick (dict, op_errstr); break; case GD_OP_LOG_FILENAME: @@ -2387,7 +2387,7 @@ glusterd_op_commit_perform (glusterd_op_t op, dict_t *dict, char **op_errstr, break; case GD_OP_REMOVE_BRICK: - ret = glusterd_op_remove_brick (dict); + ret = glusterd_op_remove_brick (dict, op_errstr); break; case GD_OP_LOG_FILENAME: @@ -2565,6 +2565,7 @@ glusterd_bricks_select_remove_brick (dict_t *dict, char **op_errstr) int32_t i = 1; char key[256] = {0,}; glusterd_pending_node_t *pending_node = NULL; + int32_t force = 0; ret = dict_get_str (dict, "volname", &volname); @@ -2586,6 +2587,12 @@ glusterd_bricks_select_remove_brick (dict_t *dict, char **op_errstr) goto out; } + ret = dict_get_int32 (dict, "force", &force); + if (ret) { + gf_log (THIS->name, GF_LOG_INFO, "force flag is not set"); + ret = 0; + goto out; + } while ( i <= count) { snprintf (key, 256, "brick%d", i); diff --git a/xlators/mgmt/glusterd/src/glusterd-rebalance.c b/xlators/mgmt/glusterd/src/glusterd-rebalance.c index a013d0adf..defaf947f 100644 --- a/xlators/mgmt/glusterd/src/glusterd-rebalance.c +++ b/xlators/mgmt/glusterd/src/glusterd-rebalance.c @@ -46,6 +46,7 @@ #include "cli1-xdr.h" #include "xdr-generic.h" +/* return values - 0: success, +ve: stopped, -ve: failure */ int gf_glusterd_rebalance_move_data (glusterd_volinfo_t *volinfo, const char *dir) { @@ -66,7 +67,8 @@ gf_glusterd_rebalance_move_data (glusterd_volinfo_t *volinfo, const char *dir) if (!fd) goto out; - if (defrag->cmd == GF_DEFRAG_CMD_START_MIGRATE_DATA_FORCE) { + if ((defrag->cmd == GF_DEFRAG_CMD_START_MIGRATE_DATA_FORCE) || + (defrag->cmd == GF_DEFRAG_CMD_START_FORCE)) { strcpy (force_string, "force"); } else { strcpy (force_string, "not-force"); @@ -105,9 +107,11 @@ gf_glusterd_rebalance_move_data (glusterd_volinfo_t *volinfo, const char *dir) } UNLOCK (&defrag->lock); - if (volinfo->defrag_status == GF_DEFRAG_STATUS_STOPED) { + if (volinfo->defrag_status != + GF_DEFRAG_STATUS_MIGRATE_DATA_STARTED) { + /* It can be one of 'stopped|paused|commit' etc */ closedir (fd); - ret = -1; + ret = 1; goto out; } } @@ -144,6 +148,7 @@ out: return ret; } +/* return values - 0: success, +ve: stopped, -ve: failure */ int gf_glusterd_rebalance_fix_layout (glusterd_volinfo_t *volinfo, const char *dir) { @@ -187,9 +192,11 @@ gf_glusterd_rebalance_fix_layout (glusterd_volinfo_t *volinfo, const char *dir) break; } - if (volinfo->defrag_status == GF_DEFRAG_STATUS_STOPED) { + if (volinfo->defrag_status != + GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED) { + /* It can be one of 'stopped|paused|commit' etc */ closedir (fd); - ret = -1; + ret = 1; goto out; } } @@ -210,6 +217,7 @@ glusterd_defrag_start (void *data) int ret = -1; struct stat stbuf = {0,}; + THIS = volinfo->xl; defrag = volinfo->defrag; if (!defrag) goto out; @@ -240,8 +248,10 @@ glusterd_defrag_start (void *data) /* Step 1: Fix layout of all the directories */ ret = gf_glusterd_rebalance_fix_layout (volinfo, defrag->mount); + if (ret < 0) + volinfo->defrag_status = GF_DEFRAG_STATUS_FAILED; + /* in both 'stopped' or 'failure' cases goto out */ if (ret) { - volinfo->defrag_status = GF_DEFRAG_STATUS_FAILED; goto out; } @@ -257,8 +267,10 @@ glusterd_defrag_start (void *data) /* Step 2: Iterate over directories to move data */ ret = gf_glusterd_rebalance_move_data (volinfo, defrag->mount); + if (ret < 0) + volinfo->defrag_status = GF_DEFRAG_STATUS_FAILED; + /* in both 'stopped' or 'failure' cases goto out */ if (ret) { - volinfo->defrag_status = GF_DEFRAG_STATUS_FAILED; goto out; } @@ -267,7 +279,8 @@ glusterd_defrag_start (void *data) } /* Completed whole process */ - if (defrag->cmd == GF_DEFRAG_CMD_START) + if ((defrag->cmd == GF_DEFRAG_CMD_START) || + (defrag->cmd == GF_DEFRAG_CMD_START_FORCE)) volinfo->defrag_status = GF_DEFRAG_STATUS_COMPLETE; volinfo->rebalance_files = defrag->total_files; @@ -281,9 +294,13 @@ out: ret = runcmd ("umount", "-l", defrag->mount, NULL); LOCK_DESTROY (&defrag->lock); + + if (defrag->cbk_fn) { + defrag->cbk_fn (volinfo, volinfo->defrag_status); + } + GF_FREE (defrag); } - return NULL; } @@ -332,7 +349,7 @@ glusterd_defrag_stop (glusterd_volinfo_t *volinfo, u_quad_t *files, LOCK (&volinfo->defrag->lock); { - volinfo->defrag_status = GF_DEFRAG_STATUS_STOPED; + volinfo->defrag_status = GF_DEFRAG_STATUS_STOPPED; *files = volinfo->defrag->total_files; *size = volinfo->defrag->total_data; } @@ -497,7 +514,7 @@ out: int glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr, - size_t len, int cmd) + size_t len, int cmd, defrag_cbk_fn_t cbk) { int ret = -1; glusterd_defrag_info_t *defrag = NULL; @@ -552,6 +569,9 @@ glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr, volinfo->defrag_status = GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED; + if (cbk) + defrag->cbk_fn = cbk; + ret = pthread_create (&defrag->th, NULL, glusterd_defrag_start, volinfo); if (ret) { @@ -635,7 +655,7 @@ glusterd_handle_defrag_volume (rpcsvc_request_t *req) case GF_DEFRAG_CMD_START_MIGRATE_DATA_FORCE: { ret = glusterd_handle_defrag_start (volinfo, msg, sizeof (msg), - cli_req.cmd); + cli_req.cmd, NULL); rsp.op_ret = ret; break; } @@ -845,7 +865,7 @@ glusterd_op_rebalance (dict_t *dict, char **op_errstr, dict_t *rsp_dict) case GF_DEFRAG_CMD_START_MIGRATE_DATA: case GF_DEFRAG_CMD_START_MIGRATE_DATA_FORCE: ret = glusterd_handle_defrag_start (volinfo, msg, sizeof (msg), - cmd); + cmd, NULL); break; case GF_DEFRAG_CMD_STOP: ret = glusterd_defrag_stop (volinfo, &files, &size, diff --git a/xlators/mgmt/glusterd/src/glusterd-replace-brick.c b/xlators/mgmt/glusterd/src/glusterd-replace-brick.c index 060d40bed..4a4289910 100644 --- a/xlators/mgmt/glusterd/src/glusterd-replace-brick.c +++ b/xlators/mgmt/glusterd/src/glusterd-replace-brick.c @@ -1486,7 +1486,7 @@ glusterd_op_perform_replace_brick (glusterd_volinfo_t *volinfo, volinfo->brick_count++; - ret = glusterd_op_perform_remove_brick (volinfo, old_brick); + ret = glusterd_op_perform_remove_brick (volinfo, old_brick, 1, NULL); if (ret) goto out; diff --git a/xlators/mgmt/glusterd/src/glusterd-store.c b/xlators/mgmt/glusterd/src/glusterd-store.c index 53fdcf484..dab075db0 100644 --- a/xlators/mgmt/glusterd/src/glusterd-store.c +++ b/xlators/mgmt/glusterd/src/glusterd-store.c @@ -321,6 +321,10 @@ glusterd_store_brickinfo_write (int fd, glusterd_brickinfo_t *brickinfo) ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_RDMA_PORT, value); + snprintf (value, sizeof(value), "%d", brickinfo->decommissioned); + ret = glusterd_store_save_value (fd, GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED, + value); + out: gf_log ("", GF_LOG_DEBUG, "Returning %d", ret); return ret; @@ -1637,6 +1641,9 @@ glusterd_store_retrieve_bricks (glusterd_volinfo_t *volinfo) pmap = pmap_registry_get (THIS); if (pmap->last_alloc <= brickinfo->rdma_port) pmap->last_alloc = brickinfo->rdma_port + 1; + } else if (!strncmp (key, GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED, + strlen (GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED))) { + gf_string2int (value, &brickinfo->decommissioned); } else { gf_log ("", GF_LOG_ERROR, "Unknown key: %s", key); diff --git a/xlators/mgmt/glusterd/src/glusterd-store.h b/xlators/mgmt/glusterd/src/glusterd-store.h index 3ca232a9a..61bda195d 100644 --- a/xlators/mgmt/glusterd/src/glusterd-store.h +++ b/xlators/mgmt/glusterd/src/glusterd-store.h @@ -63,6 +63,7 @@ typedef enum glusterd_store_ver_ac_{ #define GLUSTERD_STORE_KEY_BRICK_PATH "path" #define GLUSTERD_STORE_KEY_BRICK_PORT "listen-port" #define GLUSTERD_STORE_KEY_BRICK_RDMA_PORT "rdma.listen-port" +#define GLUSTERD_STORE_KEY_BRICK_DECOMMISSIONED "decommissioned" #define GLUSTERD_STORE_KEY_PEER_UUID "uuid" #define GLUSTERD_STORE_KEY_PEER_HOSTNAME "hostname" diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c index 8694f7536..5b247b6a9 100644 --- a/xlators/mgmt/glusterd/src/glusterd-utils.c +++ b/xlators/mgmt/glusterd/src/glusterd-utils.c @@ -548,6 +548,8 @@ glusterd_volinfo_new (glusterd_volinfo_t **volinfo) goto out; } + new_volinfo->xl = THIS; + *volinfo = new_volinfo; ret = 0; diff --git a/xlators/mgmt/glusterd/src/glusterd-volgen.c b/xlators/mgmt/glusterd/src/glusterd-volgen.c index cb198dfb9..d0533b1fc 100644 --- a/xlators/mgmt/glusterd/src/glusterd-volgen.c +++ b/xlators/mgmt/glusterd/src/glusterd-volgen.c @@ -1811,10 +1811,15 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, char **cluster_args = NULL; int i = 0; int j = 0; - int ret = 0; + int ret = -1; xlator_t *xl = NULL; xlator_t *txl = NULL; xlator_t *trav = NULL; + int removed_bricks = 0; + int index_of_removed_brick = 0; + char *removed_bricklist = NULL; + char volume_name[1024] = {0,}; + int idx = 0; volname = volinfo->volname; dict = volinfo->dict; @@ -1824,7 +1829,7 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, gf_log ("", GF_LOG_ERROR, "volume inconsistency: brick count is 0"); - return -1; + goto out; } if (volinfo->sub_count && volinfo->sub_count < volinfo->brick_count && volinfo->brick_count % volinfo->sub_count != 0) { @@ -1834,7 +1839,7 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, "number of bricks per cluster (%d) in a multi-cluster " "setup", volinfo->brick_count, volinfo->sub_count); - return -1; + goto out; } get_transport_type (volinfo, set_dict, transt, _gf_false); @@ -1844,19 +1849,32 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, i = 0; list_for_each_entry (brick, &volinfo->bricks, brick_list) { + ret = -1; xl = volgen_graph_add_nolink (graph, "protocol/client", "%s-client-%d", volname, i); if (!xl) - return -1; + goto out; ret = xlator_set_option (xl, "remote-host", brick->hostname); if (ret) - return -1; + goto out; ret = xlator_set_option (xl, "remote-subvolume", brick->path); if (ret) - return -1; + goto out; ret = xlator_set_option (xl, "transport-type", transt); if (ret) - return -1; + goto out; + if (brick->decommissioned) { + if (!removed_bricklist) { + removed_bricklist = GF_CALLOC (16 * GF_UNIT_KB, + 1, gf_common_mt_char); + index_of_removed_brick = i; + } + if (removed_bricks) + strcat (removed_bricklist, ","); + snprintf (volume_name, 1024, "%s-client-%d", volname, i); + strcat (removed_bricklist, volume_name); + removed_bricks++; + } i++; } @@ -1866,7 +1884,7 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, "differs from brick count (%d)", i, volinfo->brick_count); - return -1; + goto out; } sub_count = volinfo->sub_count; @@ -1880,15 +1898,18 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, break; case GF_CLUSTER_TYPE_STRIPE_REPLICATE: /* Replicate after the clients, then stripe */ - if (volinfo->replica_count == 0) - return -1; + if (volinfo->replica_count == 0) { + ret = -1; + goto out; + } sub_count = volinfo->replica_count; cluster_args = replicate_args; break; default: gf_log ("", GF_LOG_ERROR, "volume inconsistency: " "unrecognized clustering type"); - return -1; + ret = -1; + goto out; } i = 0; @@ -1901,14 +1922,16 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, cluster_args[0], cluster_args[1], volname, j); - if (!xl) - return -1; + if (!xl) { + ret = -1; + goto out; + } j++; } ret = volgen_xlator_link (xl, trav); if (ret) - return -1; + goto out; if (trav == txl) break; @@ -1928,14 +1951,16 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, cluster_args[0], cluster_args[1], volname, j); - if (!xl) - return -1; + if (!xl) { + ret = -1; + goto out; + } j++; } ret = volgen_xlator_link (xl, trav); if (ret) - return -1; + goto out; if (trav == txl) break; @@ -1953,8 +1978,10 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, if (dist_count > 1) { xl = volgen_graph_add_nolink (graph, "cluster/distribute", "%s-dht", volname); - if (!xl) - return -1; + if (!xl) { + ret = -1; + goto out; + } trav = xl; for (i = 0; i < dist_count; i++) @@ -1962,28 +1989,50 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, for (; trav != xl; trav = trav->prev) { ret = volgen_xlator_link (xl, trav); if (ret) - return -1; + goto out; + } + + if (removed_bricks) { + if (volinfo->sub_count) { + idx = index_of_removed_brick / volinfo->sub_count; + if (GF_CLUSTER_TYPE_REPLICATE == volinfo->type) { + snprintf (volume_name, 1024, "%s-replicate-%d", + volname, idx); + strcpy (removed_bricklist, volume_name); + } else if (volinfo->type != GF_CLUSTER_TYPE_NONE) { + snprintf (volume_name, 1024, "%s-stripe-%d ", + volname, idx); + strcpy (removed_bricklist, volume_name); + } + } + ret = xlator_set_option (xl, "decommissioned-bricks", + removed_bricklist); + if (ret) + goto out; } } ret = glusterd_volinfo_get_boolean (volinfo, VKEY_FEATURES_QUOTA); if (ret == -1) - return -1; + goto out; + if (ret) { xl = volgen_graph_add (graph, "features/quota", volname); - if (!xl) - return -1; + if (!xl) { + ret = -1; + goto out; + } } ret = volgen_graph_set_options_generic (graph, set_dict, volname, &perfxl_option_handler); if (ret) - return -1; + goto out; xl = volgen_graph_add_as (graph, "debug/io-stats", volname); if (!xl) - return -1; + goto out; ret = volgen_graph_set_options_generic (graph, set_dict, "client", &loglevel_option_handler); @@ -1991,6 +2040,11 @@ client_graph_builder (volgen_graph_t *graph, glusterd_volinfo_t *volinfo, if (!ret) ret = volgen_graph_set_options_generic (graph, set_dict, "client", &sys_loglevel_option_handler); + +out: + if (removed_bricklist) + GF_FREE (removed_bricklist); + return ret; } diff --git a/xlators/mgmt/glusterd/src/glusterd.h b/xlators/mgmt/glusterd/src/glusterd.h index 86eeaeb1c..c8fa82819 100644 --- a/xlators/mgmt/glusterd/src/glusterd.h +++ b/xlators/mgmt/glusterd/src/glusterd.h @@ -45,7 +45,7 @@ #include "glusterd1-xdr.h" #include "protocol-common.h" #include "glusterd-pmap.h" - +#include "cli1-xdr.h" #define GLUSTERD_MAX_VOLUME_NAME 1000 #define DEFAULT_LOG_FILE_DIRECTORY DATADIR "/log/glusterfs" @@ -132,6 +132,7 @@ struct glusterd_brickinfo { gf_brick_status_t status; struct rpc_clnt *rpc; gf_timer_t *timer; + int decommissioned; }; typedef struct glusterd_brickinfo glusterd_brickinfo_t; @@ -142,16 +143,11 @@ struct gf_defrag_brickinfo_ { int size; }; -typedef enum gf_defrag_status_ { - GF_DEFRAG_STATUS_NOT_STARTED, - GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED, - GF_DEFRAG_STATUS_MIGRATE_DATA_STARTED, - GF_DEFRAG_STATUS_STOPED, - GF_DEFRAG_STATUS_COMPLETE, - GF_DEFRAG_STATUS_FAILED, - GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE, - GF_DEFRAG_STATUS_MIGRATE_DATA_COMPLETE, -} gf_defrag_status_t; +struct glusterd_volinfo_; +typedef struct glusterd_volinfo_ glusterd_volinfo_t; + +typedef int (*defrag_cbk_fn_t) (glusterd_volinfo_t *volinfo, + gf_defrag_status_t status); struct glusterd_defrag_info_ { uint64_t total_files; @@ -163,6 +159,8 @@ struct glusterd_defrag_info_ { char mount[1024]; char databuf[131072]; struct gf_defrag_brickinfo_ *bricks; /* volinfo->brick_count */ + + defrag_cbk_fn_t cbk_fn; }; @@ -219,9 +217,10 @@ struct glusterd_volinfo_ { char *logdir; dict_t *gsync_slaves; -}; -typedef struct glusterd_volinfo_ glusterd_volinfo_t; + int decommission_in_progress; + xlator_t *xl; +}; typedef struct glusterd_pending_node_ { void *node; @@ -540,6 +539,8 @@ int glusterd_handle_cli_start_volume (rpcsvc_request_t *req); int glusterd_handle_cli_stop_volume (rpcsvc_request_t *req); int glusterd_handle_cli_delete_volume (rpcsvc_request_t *req); +int glusterd_handle_defrag_start (glusterd_volinfo_t *volinfo, char *op_errstr, + size_t len, int cmd, defrag_cbk_fn_t cbk); /* op-sm functions */ int glusterd_op_stage_gsync_set (dict_t *dict, char **op_errstr); @@ -565,9 +566,9 @@ int glusterd_op_stop_volume (dict_t *dict); int glusterd_op_delete_volume (dict_t *dict); int glusterd_op_add_brick (dict_t *dict, char **op_errstr); -int glusterd_op_remove_brick (dict_t *dict); +int glusterd_op_remove_brick (dict_t *dict, char **op_errstr); int glusterd_op_stage_add_brick (dict_t *dict, char **op_errstr); -int glusterd_op_stage_remove_brick (dict_t *dict); +int glusterd_op_stage_remove_brick (dict_t *dict, char **op_errstr); int glusterd_op_stage_rebalance (dict_t *dict, char **op_errstr); int glusterd_op_rebalance (dict_t *dict, char **op_errstr, dict_t *rsp_dict); @@ -575,7 +576,8 @@ int glusterd_op_rebalance (dict_t *dict, char **op_errstr, dict_t *rsp_dict); /* misc */ void glusterd_do_replace_brick (void *data); -int glusterd_op_perform_remove_brick (glusterd_volinfo_t *volinfo, char *brick); +int glusterd_op_perform_remove_brick (glusterd_volinfo_t *volinfo, char *brick, + int force, int *need_migrate); int glusterd_op_stop_volume_args_get (dict_t *dict, char** volname, int *flags); -- cgit