Self-heald: Enable configuring of heal poll timeout

Change-Id: I631e5bf4b3615b553b72e7ac7f490714b3b995f9
BUG: 821395
Signed-off-by: Pranith Kumar K <pranithk@gluster.com>
Reviewed-on: http://review.gluster.com/3329
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>

4 files changed, 74 insertions, 8 deletions

diff --git a/xlators/cluster/afr/src/afr-self-heald.c b/xlators/cluster/afr/src/afr-self-heald.c
index 3a6295749b5..1a749249056 100644
--- a/xlators/cluster/afr/src/afr-self-heald.c
+++ b/xlators/cluster/afr/src/afr-self-heald.c
@@ -19,8 +19,6 @@
 #include "protocol-common.h"
 #include "event-history.h"
 
-#define AFR_POLL_TIMEOUT 600
-
 typedef enum {
         STOP_CRAWL_ON_SINGLE_SUBVOL = 1
 } afr_crawl_flags_t;
@@ -602,7 +600,7 @@ afr_poll_self_heal (void *data)
         }
         if (shd->enabled && (shd->pos[child] == AFR_POS_LOCAL))
                 _do_self_heal_on_subvol (this, child, INDEX);
-        timeout.tv_sec = AFR_POLL_TIMEOUT;
+        timeout.tv_sec = shd->timeout;
         timeout.tv_usec = 0;
         //notify and previous timer should be synchronized.
         LOCK (&priv->lock);
diff --git a/xlators/cluster/afr/src/afr.c b/xlators/cluster/afr/src/afr.c
index 3f102f80885..8e94d549737 100644
--- a/xlators/cluster/afr/src/afr.c
+++ b/xlators/cluster/afr/src/afr.c
@@ -164,6 +164,8 @@ reconfigure (xlator_t *this, dict_t *options)
         GF_OPTION_RECONF ("quorum-count", priv->quorum_count, options,
                           uint32, out);
         fix_quorum_options(this,priv,qtype);
+        GF_OPTION_RECONF ("heal-timeout", priv->shd.timeout, options,
+                          int32, out);
 
         ret = 0;
 out:
@@ -402,6 +404,7 @@ init (xlator_t *this)
                 goto out;
         priv->root_inode = inode_ref (this->itable->root);
         GF_OPTION_INIT ("node-uuid", priv->shd.node_uuid, str, out);
+        GF_OPTION_INIT ("heal-timeout", priv->shd.timeout, int32, out);
 
         ret = 0;
 out:
@@ -596,5 +599,12 @@ struct volume_options options[] = {
           .type = GF_OPTION_TYPE_STR,
           .description = "Local glusterd uuid string",
         },
+        { .key  = {"heal-timeout"},
+          .type = GF_OPTION_TYPE_INT,
+          .min  = 60,
+          .max  = INT_MAX,
+          .default_value = "600",
+          .description = "Poll timeout for checking the need to self-heal"
+        },
         { .key  = {NULL} },
 };
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index 497d82d69ce..f00a8498db4 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -96,6 +96,7 @@ typedef struct afr_self_heald_ {
         eh_t             *heal_failed;
         eh_t             *split_brain;
         char             *node_uuid;
+        int              timeout;
 } afr_self_heald_t;
 
 typedef struct _afr_private {
diff --git a/xlators/mgmt/glusterd/src/glusterd-volgen.c b/xlators/mgmt/glusterd/src/glusterd-volgen.c
index 043f6146b21..1d5241798bd 100644
--- a/xlators/mgmt/glusterd/src/glusterd-volgen.c
+++ b/xlators/mgmt/glusterd/src/glusterd-volgen.c
@@ -125,6 +125,7 @@ static struct volopt_map_entry glusterd_volopt_map[] = {
         {"cluster.data-self-heal",               "cluster/replicate",  NULL, NULL, NO_DOC, 0     },
         {"cluster.entry-self-heal",              "cluster/replicate",  NULL, NULL, NO_DOC, 0     },
         {"cluster.self-heal-daemon",             "cluster/replicate",  "!self-heal-daemon" , NULL, NO_DOC, 0     },
+        {"cluster.heal-timeout",                 "cluster/replicate",  "!heal-timeout" , NULL, NO_DOC, 0     },
         {"cluster.strict-readdir",               "cluster/replicate",  NULL, NULL, NO_DOC, 0     },
         {"cluster.self-heal-window-size",        "cluster/replicate",         "data-self-heal-window-size", NULL, DOC, 0},
         {"cluster.data-change-log",              "cluster/replicate",  NULL, NULL, NO_DOC, 0     },
@@ -2535,20 +2536,35 @@ build_client_graph (volgen_graph_t *graph, glusterd_volinfo_t *volinfo,
                                     &client_graph_builder);
 }
 
+char *gd_shd_options[] = {
+        "!self-heal-daemon",
+        "!heal-timeout",
+        NULL
+};
+
+char*
+gd_get_matching_option (char **options, char *option)
+{
+        while (*options && strcmp (*options, option))
+                options++;
+        return *options;
+}
+
 static int
 shd_option_handler (volgen_graph_t *graph, struct volopt_map_entry *vme,
                     void *param)
 {
         int                     ret = 0;
         struct volopt_map_entry new_vme = {0};
-        int                     shd = 0;
+        char                    *shd_option = NULL;
 
-        shd = !strcmp (vme->option, "!self-heal-daemon");
-        if ((vme->option[0] == '!') && !shd)
+        if (vme->option[0] != '!')
+                goto out;
+        shd_option = gd_get_matching_option (gd_shd_options, vme->option);
+        if (!shd_option)
                 goto out;
         new_vme = *vme;
-        if (shd)
-                new_vme.option = "self-heal-daemon";
+        new_vme.option = shd_option + 1;//option with out '!'
 
         ret = no_filter_option_handler (graph, &new_vme, param);
 out:
@@ -3415,6 +3431,26 @@ glusterd_delete_volfile (glusterd_volinfo_t *volinfo,
 }
 
 int
+validate_shdopts (glusterd_volinfo_t *volinfo,
+                  dict_t *val_dict,
+                  char **op_errstr)
+{
+        volgen_graph_t graph = {0,};
+        int     ret = -1;
+
+        graph.errstr = op_errstr;
+
+        ret = build_shd_graph (&graph, val_dict);
+        if (!ret)
+                ret = graph_reconf_validateopt (&graph.graph, op_errstr);
+
+        volgen_graph_free (&graph);
+
+        gf_log ("glusterd", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
+
+int
 validate_nfsopts (glusterd_volinfo_t *volinfo,
                     dict_t *val_dict,
                     char **op_errstr)
@@ -3618,6 +3654,16 @@ glusterd_validate_globalopts (glusterd_volinfo_t *volinfo,
         }
 
         ret = validate_nfsopts (volinfo, val_dict, op_errstr);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Could not Validate nfs");
+                goto out;
+        }
+
+        ret = validate_shdopts (volinfo, val_dict, op_errstr);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Could not Validate self-heald");
+                goto out;
+        }
 
 out:
         gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
@@ -3667,6 +3713,17 @@ glusterd_validate_reconfopts (glusterd_volinfo_t *volinfo, dict_t *val_dict,
         }
 
         ret = validate_nfsopts (volinfo, val_dict, op_errstr);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Could not Validate nfs");
+                goto out;
+        }
+
+
+        ret = validate_shdopts (volinfo, val_dict, op_errstr);
+        if (ret) {
+                gf_log ("", GF_LOG_DEBUG, "Could not Validate self-heald");
+                goto out;
+        }
 
 
 out: