cluster/afr: Add dry-run functionality to self-heal.

This will be useful in figuring out if a file needs
self-heal or not with certainity for data-self-heal

Change-Id: Idf98a68e69f2c35646ef2e7c97302586fe1dc07d
BUG: 1039544
Signed-off-by: Pranith Kumar K <pkarampu@redhat.com>
Reviewed-on: http://review.gluster.org/6510
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Ravishankar N <ravishankar@redhat.com>
Reviewed-by: Vijay Bellur <vbellur@redhat.com>

6 files changed, 40 insertions, 7 deletions

diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index 66978a910aa..953df5a67a4 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -1743,14 +1743,23 @@ afr_self_heal_lookup_unwind (call_frame_t *frame, xlator_t *this,
 
         afr_lookup_done_success_action (frame, this, _gf_true);
         xattr = local->cont.lookup.xattr;
-        if (xattr) {
+        if (!xattr)
+                goto out;
+
+        if (sh_failed) {
                 ret = dict_set_int32 (xattr, "sh-failed", sh_failed);
                 if (ret)
                         gf_log (this->name, GF_LOG_ERROR, "%s: Failed to set "
                                 "sh-failed to %d", local->loc.path, sh_failed);
-
-                if (local->self_heal.actual_sh_started == _gf_true &&
-                    sh_failed == 0) {
+        } else {
+                ret = dict_set_int32 (xattr, "metadata-self-heal-pending",
+                                      local->self_heal.metadata_sh_pending);
+                ret = dict_set_int32 (xattr, "data-self-heal-pending",
+                                      local->self_heal.data_sh_pending);
+                ret = dict_set_int32 (xattr, "entry-self-heal-pending",
+                                      local->self_heal.entry_sh_pending);
+
+                if (local->self_heal.actual_sh_started == _gf_true) {
                         ret = dict_set_int32 (xattr, "actual-sh-done", 1);
                         if (ret)
                                 gf_log(this->name, GF_LOG_ERROR, "%s: Failed to"
@@ -2511,6 +2520,10 @@ afr_lookup (call_frame_t *frame, xlator_t *this,
                               &local->foreground_self_heal);
         dict_del (xattr_req, "foreground-self-heal");
 
+        ret = dict_get_int32 (xattr_req, "dry-run-self-heal",
+                              &local->self_heal.dry_run);
+        dict_del (xattr_req, "dry-run-self-heal");
+
         ret = afr_lookup_xattr_req_prepare (local, this, xattr_req, &local->loc,
                                             &gfid_req);
         if (ret) {
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c
index df537086e62..6f82761b33d 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
@@ -2260,6 +2260,7 @@ afr_self_heal_local_init (afr_local_t *l, xlator_t *this)
 
         shc->unwind = sh->unwind;
         shc->gfid_sh_success_cbk = sh->gfid_sh_success_cbk;
+        shc->dry_run = sh->dry_run;
         shc->do_missing_entry_self_heal = sh->do_missing_entry_self_heal;
         shc->do_gfid_self_heal = sh->do_gfid_self_heal;
         shc->do_data_self_heal = sh->do_data_self_heal;
@@ -2352,6 +2353,9 @@ afr_self_heal_completion_cbk (call_frame_t *bgsh_frame, xlator_t *this)
                 orig_frame_local = sh->orig_frame->local;
                 orig_frame_sh = &orig_frame_local->self_heal;
                 orig_frame_sh->actual_sh_started = _gf_true;
+                orig_frame_sh->entry_sh_pending = sh->entry_sh_pending;
+                orig_frame_sh->data_sh_pending = sh->data_sh_pending;
+                orig_frame_sh->metadata_sh_pending = sh->metadata_sh_pending;
                 sh->unwind (sh->orig_frame, this, sh->op_ret, sh->op_errno,
                             is_self_heal_failed (sh, AFR_CHECK_ALL));
         }
diff --git a/xlators/cluster/afr/src/afr-self-heal-data.c b/xlators/cluster/afr/src/afr-self-heal-data.c
index 9de26ee569c..e5ca84292fa 100644
--- a/xlators/cluster/afr/src/afr-self-heal-data.c
+++ b/xlators/cluster/afr/src/afr-self-heal-data.c
@@ -1042,6 +1042,7 @@ afr_sh_data_fxattrop_fstat_done (call_frame_t *frame, xlator_t *this)
                 afr_sh_data_setattr (frame, this, &sh->buf[tstamp_source]);
         } else {
                 afr_set_data_sh_info_str (local, sh, this);
+
                 if (nsources == 0) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "No self-heal needed for %s",
@@ -1051,7 +1052,9 @@ afr_sh_data_fxattrop_fstat_done (call_frame_t *frame, xlator_t *this)
                         return 0;
                 }
 
-                if (sh->do_data_self_heal &&
+                sh->data_sh_pending = _gf_true;
+
+                if (!sh->dry_run && sh->do_data_self_heal &&
                     afr_data_self_heal_enabled (priv->data_self_heal))
                         afr_sh_data_fix (frame, this);
                 else
diff --git a/xlators/cluster/afr/src/afr-self-heal-entry.c b/xlators/cluster/afr/src/afr-self-heal-entry.c
index 0ca06aaa382..8359079cea4 100644
--- a/xlators/cluster/afr/src/afr-self-heal-entry.c
+++ b/xlators/cluster/afr/src/afr-self-heal-entry.c
@@ -2343,7 +2343,12 @@ afr_sh_entry_fix (call_frame_t *frame, xlator_t *this,
                                         sh->fresh_children);
 
 heal:
-        afr_sh_entry_sync_prepare (frame, this);
+        sh->entry_sh_pending = _gf_true;
+        if (sh->dry_run) {
+                afr_sh_entry_finish (frame, this);
+        } else {
+                afr_sh_entry_sync_prepare (frame, this);
+        }
 out:
         return;
 }
diff --git a/xlators/cluster/afr/src/afr-self-heal-metadata.c b/xlators/cluster/afr/src/afr-self-heal-metadata.c
index fd5da6cfd33..e26d3580e51 100644
--- a/xlators/cluster/afr/src/afr-self-heal-metadata.c
+++ b/xlators/cluster/afr/src/afr-self-heal-metadata.c
@@ -672,7 +672,9 @@ afr_sh_metadata_fix (call_frame_t *frame, xlator_t *this,
                                         sh->fresh_children);
         }
 
-        if (sh->do_metadata_self_heal && priv->metadata_self_heal)
+        sh->metadata_sh_pending = _gf_true;
+        if (!sh->dry_run &&
+            sh->do_metadata_self_heal && priv->metadata_self_heal)
                 afr_sh_metadata_sync_prepare (frame, this);
         else
                 afr_sh_metadata_finish (frame, this);
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index dc0bc06fb70..676bd9b9466 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -285,6 +285,12 @@ struct afr_self_heal_ {
         gf_boolean_t entries_skipped;
 
         gf_boolean_t actual_sh_started;
+
+        int32_t dry_run;
+        gf_boolean_t metadata_sh_pending;
+        gf_boolean_t data_sh_pending;
+        gf_boolean_t entry_sh_pending;
+
         gf_boolean_t sync_done;
         gf_boolean_t data_lock_held;
         gf_boolean_t sh_dom_lock_held;