Self-heald: Handle errors gracefully and show errors to users

Change-Id: I5424ebfadb5b2773ee6f7370cc2867a555aa48dd
BUG: 800352
Signed-off-by: Pranith Kumar K <pranithk@gluster.com>
Reviewed-on: http://review.gluster.com/2962
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>

6 files changed, 64 insertions, 33 deletions

diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.c b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
index d96d6717eeb..ce9581644b1 100644
--- a/xlators/mgmt/glusterd/src/glusterd-op-sm.c
+++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.c
@@ -2780,9 +2780,9 @@ _heal_volume_add_shd_rsp (dict_t *this, char *key, data_t *value, void *data)
         int                             rxl_child_id = 0;
         int                             brick_id = 0;
         int                             int_len = 0;
-        int                             brick_count = 0;
         int                             ret = 0;
         glusterd_heal_rsp_conv_t        *rsp_ctx = NULL;
+        glusterd_brickinfo_t            *brickinfo = NULL;
 
         rsp_ctx = data;
         rxl_end = strchr (key, '-');
@@ -2810,13 +2810,19 @@ _heal_volume_add_shd_rsp (dict_t *this, char *key, data_t *value, void *data)
         volinfo = rsp_ctx->volinfo;
         brick_id = rxl_id * volinfo->replica_count + rxl_child_id;
 
+        if (!strcmp (rxl_child_end, "status")) {
+                brickinfo = glusterd_get_brickinfo_by_position (volinfo,
+                                                                brick_id);
+                if (!brickinfo)
+                        goto out;
+                if (!glusterd_is_local_brick (rsp_ctx->this, volinfo,
+                                              brickinfo))
+                        goto out;
+        }
         new_value = data_copy (value);
         snprintf (new_key, sizeof (new_key), "%d%s", brick_id, rxl_child_end);
         dict_set (rsp_ctx->dict, new_key, new_value);
 
-        ret = dict_get_int32 (rsp_ctx->dict, "count", &brick_count);
-        if (brick_id >= brick_count)
-                ret = dict_set_int32 (rsp_ctx->dict, "count", brick_id + 1);
 out:
         return;
 }
@@ -2847,6 +2853,7 @@ glusterd_heal_volume_brick_rsp (dict_t *req_dict, dict_t *rsp_dict,
 
         rsp_ctx.dict = op_ctx;
         rsp_ctx.volinfo = volinfo;
+        rsp_ctx.this = THIS;
         dict_foreach (rsp_dict, _heal_volume_add_shd_rsp, &rsp_ctx);
 
 out:
diff --git a/xlators/mgmt/glusterd/src/glusterd-op-sm.h b/xlators/mgmt/glusterd/src/glusterd-op-sm.h
index bfc41b29252..1f32681a5a1 100644
--- a/xlators/mgmt/glusterd/src/glusterd-op-sm.h
+++ b/xlators/mgmt/glusterd/src/glusterd-op-sm.h
@@ -157,6 +157,7 @@ typedef struct glusterd_pr_brick_rsp_conv_t {
 typedef struct glusterd_heal_rsp_conv_ {
         dict_t *dict;
         glusterd_volinfo_t *volinfo;
+        xlator_t *this;
 } glusterd_heal_rsp_conv_t;
 
 typedef struct glusterd_status_rsp_conv_ {
diff --git a/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c b/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c
index c7931dbfafc..649156f4b33 100644
--- a/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c
+++ b/xlators/mgmt/glusterd/src/glusterd-rpc-ops.c
@@ -111,11 +111,6 @@ glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret,
                 }
                 break;
         }
-        case GD_OP_HEAL_VOLUME:
-        {
-                glusterd_add_bricks_hname_path_to_dict (ctx);
-                break;
-        }
         case GD_OP_PROFILE_VOLUME:
         {
                 if (ctx && dict_get_int32 (ctx, "count", &count)) {
@@ -153,6 +148,7 @@ glusterd_op_send_cli_response (glusterd_op_t op, int32_t op_ret,
         case GD_OP_SET_VOLUME:
         case GD_OP_LIST_VOLUME:
         case GD_OP_CLEARLOCKS_VOLUME:
+        case GD_OP_HEAL_VOLUME:
         {
                 /*nothing specific to be done*/
                 break;
@@ -1198,28 +1194,6 @@ out:
         return ret;
 }
 
-void
-_heal_volume_add_peer_rsp (dict_t *peer_dict, char *key, data_t *value,
-                           void *data)
-{
-        int                             max_brick = 0;
-        int                             peer_max_brick = 0;
-        int                             ret = 0;
-        dict_t                          *ctx_dict = data;
-
-
-
-        ret = dict_get_int32 (ctx_dict, "count", &max_brick);
-        ret = dict_get_int32 (peer_dict, "count", &peer_max_brick);
-        if (peer_max_brick > max_brick)
-                ret = dict_set_int32 (ctx_dict, "count", peer_max_brick);
-        else
-                ret = dict_set_int32 (ctx_dict, "count", max_brick);
-        dict_del (peer_dict, "count");
-        dict_copy (peer_dict, ctx_dict);
-        return;
-}
-
 int
 glusterd_volume_heal_use_rsp_dict (dict_t *rsp_dict)
 {
@@ -1236,7 +1210,7 @@ glusterd_volume_heal_use_rsp_dict (dict_t *rsp_dict)
 
         if (!ctx_dict)
                 goto out;
-        dict_foreach (rsp_dict, _heal_volume_add_peer_rsp, ctx_dict);
+        dict_copy (rsp_dict, ctx_dict);
 out:
         return ret;
 }
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index 9cc436496ab..4b84039dfe6 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -3303,9 +3303,23 @@ glusterd_get_brickinfo (xlator_t *this, const char *brickname, int port,
         return ret;
 }
 
+glusterd_brickinfo_t*
+glusterd_get_brickinfo_by_position (glusterd_volinfo_t *volinfo, uint32_t pos)
+{
+        glusterd_brickinfo_t    *tmpbrkinfo = NULL;
+
+        list_for_each_entry (tmpbrkinfo, &volinfo->bricks,
+                             brick_list) {
+                if (pos == 0)
+                        return tmpbrkinfo;
+                pos--;
+        }
+        return NULL;
+}
+
 void
 glusterd_set_brick_status (glusterd_brickinfo_t  *brickinfo,
-                            gf_brick_status_t status)
+                           gf_brick_status_t status)
 {
         GF_ASSERT (brickinfo);
         brickinfo->status = status;
@@ -5139,3 +5153,21 @@ glusterd_uuid_to_hostname (uuid_t uuid)
         return hostname;
 }
 
+gf_boolean_t
+glusterd_is_local_brick (xlator_t *this, glusterd_volinfo_t *volinfo,
+                         glusterd_brickinfo_t *brickinfo)
+{
+        gf_boolean_t    local = _gf_false;
+        int             ret = 0;
+        glusterd_conf_t *conf = NULL;
+
+        if (uuid_is_null (brickinfo->uuid)) {
+                ret = glusterd_resolve_brick (brickinfo);
+                if (ret)
+                        goto out;
+        }
+        conf = this->private;
+        local = !uuid_compare (brickinfo->uuid, conf->uuid);
+out:
+        return local;
+}
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.h b/xlators/mgmt/glusterd/src/glusterd-utils.h
index 7b5a387c275..fa9f7737005 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.h
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.h
@@ -419,4 +419,11 @@ glusterd_add_node_to_dict (char *server, dict_t *dict, int count);
 
 char *
 glusterd_uuid_to_hostname (uuid_t uuid);
+
+glusterd_brickinfo_t*
+glusterd_get_brickinfo_by_position (glusterd_volinfo_t *volinfo, uint32_t pos);
+
+gf_boolean_t
+glusterd_is_local_brick (xlator_t *this, glusterd_volinfo_t *volinfo,
+                         glusterd_brickinfo_t *brickinfo);
 #endif
diff --git a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
index d6e58c1b844..93c00983a21 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volume-ops.c
@@ -448,6 +448,7 @@ glusterd_handle_cli_heal_volume (rpcsvc_request_t *req)
         dict_t                          *dict = NULL;
         glusterd_op_t                   cli_op = GD_OP_HEAL_VOLUME;
         char                            *volname = NULL;
+        glusterd_volinfo_t              *volinfo = NULL;
 
         GF_ASSERT (req);
 
@@ -483,6 +484,15 @@ glusterd_handle_cli_heal_volume (rpcsvc_request_t *req)
         gf_log ("glusterd", GF_LOG_INFO, "Received heal vol req"
                 "for volume %s", volname);
 
+        ret = glusterd_add_bricks_hname_path_to_dict (dict);
+        if (ret)
+                goto out;
+        ret = glusterd_volinfo_find (volname, &volinfo);
+        if (ret)
+                goto out;
+        ret = dict_set_int32 (dict, "count", volinfo->brick_count);
+        if (ret)
+                goto out;
         ret = glusterd_op_begin (req, GD_OP_HEAL_VOLUME, dict);
 
         gf_cmd_log ("volume heal","on volname: %s %s", volname,