cluster/dht: Add "afr.readdir-failover=off" option the rebalance process

By failing over readdir (default behaviour), rebalance could get duplicate
files, as readdir would re-read from offset 0. Rebalance should not attempt
to migrate these files again.

Additionally, we need to handle these cases as failure in rebalance crawl.

No test case provided, as we cannot determine the read child in afr.

Change-Id: If07508b4f92dacc17e0f695b48a866c7c66004be
BUG: 859387
Signed-off-by: shishir gowda <sgowda@redhat.com>
Reviewed-on: http://review.gluster.org/4300
Tested-by: Gluster Build System <jenkins@build.gluster.com>
Reviewed-by: Anand Avati <avati@redhat.com>

1 files changed, 27 insertions, 7 deletions

diff --git a/xlators/cluster/dht/src/dht-rebalance.c b/xlators/cluster/dht/src/dht-rebalance.c
index a2e96a1b281..bc5252ee9cd 100644
--- a/xlators/cluster/dht/src/dht-rebalance.c
+++ b/xlators/cluster/dht/src/dht-rebalance.c
@@ -1080,17 +1080,24 @@ gf_defrag_migrate_data (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
 
         while ((ret = syncop_readdirp (this, fd, 131072, offset, NULL,
                                        &entries)) != 0) {
-                if (ret < 0)
-                        break;
+
+                if (ret < 0) {
+
+                        gf_log (this->name, GF_LOG_ERROR, "Readdir returned %s."
+                                " Aborting migrate-data",
+                                strerror(readdir_operrno));
+                        goto out;
+                }
 
                 /* Need to keep track of ENOENT errno, that means, there is no
                    need to send more readdirp() */
                 readdir_operrno = errno;
 
-                free_entries = _gf_true;
-
                 if (list_empty (&entries.list))
                         break;
+
+                free_entries = _gf_true;
+
                 list_for_each_entry_safe (entry, tmp, &entries.list, list) {
                         if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
                                 ret = 1;
@@ -1284,6 +1291,7 @@ gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
         dict_t                  *dict           = NULL;
         off_t                    offset         = 0;
         struct iatt              iatt           = {0,};
+        int                      readdirp_errno = 0;
 
         ret = syncop_lookup (this, loc, NULL, &iatt, NULL, NULL);
         if (ret) {
@@ -1319,12 +1327,22 @@ gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
         while ((ret = syncop_readdirp (this, fd, 131072, offset, NULL,
                 &entries)) != 0)
         {
-                if ((ret < 0) || (ret && (errno == ENOENT)))
-                        break;
-                free_entries = _gf_true;
+
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "Readdir returned %s"
+                                ". Aborting fix-layout",strerror(errno));
+                        goto out;
+                }
+
+                /* Need to keep track of ENOENT errno, that means, there is no
+                   need to send more readdirp() */
+                readdirp_errno = errno;
 
                 if (list_empty (&entries.list))
                         break;
+
+                free_entries = _gf_true;
+
                 list_for_each_entry_safe (entry, tmp, &entries.list, list) {
                         if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
                                 ret = 1;
@@ -1400,6 +1418,8 @@ gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
                 gf_dirent_free (&entries);
                 free_entries = _gf_false;
                 INIT_LIST_HEAD (&entries.list);
+                if (readdirp_errno == ENOENT)
+                        break;
         }
 
         ret = 0;