summaryrefslogtreecommitdiffstats
path: root/xlators/mgmt/glusterd/src/glusterd-utils.c
diff options
context:
space:
mode:
authorAvra Sengupta <asengupt@redhat.com>2014-04-22 10:27:23 +0000
committerKrishnan Parthasarathi <kparthas@redhat.com>2014-05-06 00:25:11 -0700
commit86fccad56d2ffd6c65e1571ff87dbf625b8ce55e (patch)
treef78322f5b0bad5457e34935b336090fe3eb5d60b /xlators/mgmt/glusterd/src/glusterd-utils.c
parentf846e54b8844decbc8bd73840e7d35b2dcaed2e0 (diff)
glusterd/snashot: Perform missed snap creates
When a brick is started, and the glusterfsd process requests for volfile, the brick_name is sent in the req dict. In glusterd, after fetching the spec the brick_name is looked up in the missed_snap_list, and any missing snap creates on the same brick are performed. After this, the glusterd responds back with the specfile. Also collate brick data from the node's hosting the bricks during restore. In case the data is absent, the local node's data is used. This is needed to ensure that, during a restore we collect the information created when a missed snap create is performed. Change-Id: I47cefdeba96f2702be810965734cf0fac61d3d2d BUG: 1061685 Signed-off-by: Avra Sengupta <asengupt@redhat.com> Reviewed-on: http://review.gluster.org/7551 Reviewed-by: Santosh Pradhan <spradhan@redhat.com> Reviewed-by: Rajesh Joseph <rjoseph@redhat.com> Tested-by: Gluster Build System <jenkins@build.gluster.com> Reviewed-by: Krishnan Parthasarathi <kparthas@redhat.com> Tested-by: Krishnan Parthasarathi <kparthas@redhat.com>
Diffstat (limited to 'xlators/mgmt/glusterd/src/glusterd-utils.c')
-rw-r--r--xlators/mgmt/glusterd/src/glusterd-utils.c118
1 files changed, 76 insertions, 42 deletions
diff --git a/xlators/mgmt/glusterd/src/glusterd-utils.c b/xlators/mgmt/glusterd/src/glusterd-utils.c
index d508e74f510..b955fd9de9d 100644
--- a/xlators/mgmt/glusterd/src/glusterd-utils.c
+++ b/xlators/mgmt/glusterd/src/glusterd-utils.c
@@ -647,10 +647,13 @@ out:
* TODO: Duplicate all members of volinfo, e.g. geo-rep sync slaves
*/
int32_t
-glusterd_snap_volinfo_restore (dict_t *rsp_dict,
+glusterd_snap_volinfo_restore (dict_t *dict, dict_t *rsp_dict,
glusterd_volinfo_t *new_volinfo,
- glusterd_volinfo_t *snap_volinfo)
+ glusterd_volinfo_t *snap_volinfo,
+ int32_t volcount)
{
+ char *value = NULL;
+ char key[PATH_MAX] = "";
int32_t brick_count = -1;
int32_t ret = -1;
xlator_t *this = NULL;
@@ -659,6 +662,7 @@ glusterd_snap_volinfo_restore (dict_t *rsp_dict,
this = THIS;
GF_ASSERT (this);
+ GF_ASSERT (dict);
GF_ASSERT (rsp_dict);
GF_VALIDATE_OR_GOTO (this->name, new_volinfo, out);
@@ -666,6 +670,7 @@ glusterd_snap_volinfo_restore (dict_t *rsp_dict,
brick_count = 0;
list_for_each_entry (brickinfo, &snap_volinfo->bricks, brick_list) {
+ brick_count++;
ret = glusterd_brickinfo_new (&new_brickinfo);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "Failed to create "
@@ -681,6 +686,28 @@ glusterd_snap_volinfo_restore (dict_t *rsp_dict,
goto out;
}
+ /* Fetch values if present in dict These values won't
+ * be present in case of a missed restore. In that case
+ * it's fine to use the local node's value
+ */
+ snprintf (key, sizeof (key), "snap%d.brick%d.path",
+ volcount, brick_count);
+ ret = dict_get_str (dict, key, &value);
+ if (!ret)
+ strncpy (new_brickinfo->path, value,
+ sizeof(new_brickinfo->path));
+
+ snprintf (key, sizeof (key), "snap%d.brick%d.snap_status",
+ volcount, brick_count);
+ ret = dict_get_int32 (dict, key, &new_brickinfo->snap_status);
+
+ snprintf (key, sizeof (key), "snap%d.brick%d.device_path",
+ volcount, brick_count);
+ ret = dict_get_str (dict, key, &value);
+ if (!ret)
+ strncpy (new_brickinfo->device_path, value,
+ sizeof(new_brickinfo->device_path));
+
/* If the brick is not of this peer, or snapshot is missed *
* for the brick do not replace the xattr for it */
if ((!uuid_compare (brickinfo->uuid, MY_UUID)) &&
@@ -713,7 +740,7 @@ glusterd_snap_volinfo_restore (dict_t *rsp_dict,
(rsp_dict,
snap_volinfo,
brickinfo,
- brick_count + 1,
+ brick_count,
GF_SNAP_OPTION_TYPE_RESTORE);
if (ret) {
gf_log (this->name, GF_LOG_ERROR,
@@ -729,7 +756,6 @@ glusterd_snap_volinfo_restore (dict_t *rsp_dict,
&new_volinfo->bricks);
/* ownership of new_brickinfo is passed to new_volinfo */
new_brickinfo = NULL;
- brick_count++;
}
/* Regenerate all volfiles */
@@ -2336,7 +2362,8 @@ glusterd_add_volume_to_dict (glusterd_volinfo_t *volinfo,
if (ret)
goto out;
- snprintf (key, sizeof (key), "volume%d.restored_from_snap", count);
+ snprintf (key, sizeof (key), "%s%d.restored_from_snap",
+ prefix, count);
ret = dict_set_dynstr_with_alloc
(dict, key,
uuid_utoa (volinfo->restored_from_snap));
@@ -4562,6 +4589,7 @@ glusterd_perform_missed_op (glusterd_snap_t *snap, int32_t op)
glusterd_conf_t *priv = NULL;
glusterd_volinfo_t *snap_volinfo = NULL;
glusterd_volinfo_t *volinfo = NULL;
+ glusterd_volinfo_t *tmp = NULL;
xlator_t *this = NULL;
uuid_t null_uuid = {0};
@@ -4590,35 +4618,37 @@ glusterd_perform_missed_op (glusterd_snap_t *snap, int32_t op)
break;
case GF_SNAP_OPTION_TYPE_RESTORE:
- /* TODO : As of now there is only volume in snapshot.
- * Change this when multiple volume snapshot is introduced
- */
- snap_volinfo = list_entry (snap->volumes.next,
- glusterd_volinfo_t, vol_list);
-
- /* Find the parent volinfo */
- ret = glusterd_volinfo_find (snap_volinfo->parent_volname,
- &volinfo);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR,
- "Could not get volinfo of %s",
- snap_volinfo->parent_volname);
- goto out;
- }
+ list_for_each_entry_safe (snap_volinfo, tmp,
+ &snap->volumes, vol_list) {
+ ret = glusterd_volinfo_find
+ (snap_volinfo->parent_volname,
+ &volinfo);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "Could not get volinfo of %s",
+ snap_volinfo->parent_volname);
+ goto out;
+ }
- /* Bump down the original volinfo's version, coz it would have
- * incremented already due to volume handshake
- */
- volinfo->version--;
- uuid_copy (volinfo->restored_from_snap, null_uuid);
+ volinfo->version--;
+ uuid_copy (volinfo->restored_from_snap, null_uuid);
- /* Perform the restore */
- ret = gd_restore_snap_volume (dict, volinfo, snap_volinfo);
- if (ret) {
- gf_log (this->name, GF_LOG_ERROR, "Failed to restore "
- "snap for %s", snap->snapname);
- volinfo->version++;
- goto out;
+ /* gd_restore_snap_volume() uses the dict and volcount
+ * to fetch snap brick info from other nodes, which were
+ * collected during prevalidation. As this is an ad-hoc
+ * op and only local node's data matter, hence sending
+ * volcount as 0 and re-using the same dict because we
+ * need not record any missed creates in the rsp_dict.
+ */
+ ret = gd_restore_snap_volume (dict, dict, volinfo,
+ snap_volinfo, 0);
+ if (ret) {
+ gf_log (this->name, GF_LOG_ERROR,
+ "Failed to restore snap for %s",
+ snap->snapname);
+ volinfo->version++;
+ goto out;
+ }
}
break;
@@ -4851,7 +4881,9 @@ out:
return is_local;
}
-/* Check if the peer has missed any snap delete for the given snap_id */
+/* Check if the peer has missed any snap delete
+ * or restore for the given snap_id
+ */
gf_boolean_t
glusterd_peer_has_missed_snap_delete (glusterd_peerinfo_t *peerinfo,
char *peer_snap_id)
@@ -4885,8 +4917,10 @@ glusterd_peer_has_missed_snap_delete (glusterd_peerinfo_t *peerinfo,
list_for_each_entry (snap_opinfo,
&missed_snapinfo->snap_ops,
snap_ops_list) {
- if ((snap_opinfo->op ==
- GF_SNAP_OPTION_TYPE_DELETE) &&
+ if (((snap_opinfo->op ==
+ GF_SNAP_OPTION_TYPE_DELETE) ||
+ (snap_opinfo->op ==
+ GF_SNAP_OPTION_TYPE_RESTORE)) &&
(snap_opinfo->status ==
GD_MISSED_SNAP_PENDING)) {
missed_delete = _gf_true;
@@ -5141,7 +5175,7 @@ out:
* glusterd_compare_and_update_snap() implements the following algorithm to
* perform the above task:
* Step 1: Start.
- * Step 2: Check if the peer is missing a delete on the said snap.
+ * Step 2: Check if the peer is missing a delete or restore on the said snap.
* If yes, goto step 6.
* Step 3: Check if there is a conflict between the peer's data and the
* local snap. If no, goto step 5.
@@ -5205,8 +5239,8 @@ glusterd_compare_and_update_snap (dict_t *peer_data, int32_t snap_count,
goto out;
}
- /* Check if the peer has missed a snap delete for the
- * snap in question
+ /* Check if the peer has missed a snap delete or restore
+ * resulting in stale data for the snap in question
*/
missed_delete = glusterd_peer_has_missed_snap_delete (peerinfo,
peer_snap_id);
@@ -6983,7 +7017,7 @@ glusterd_add_brick_mount_details (glusterd_brickinfo_t *brickinfo,
}
char*
-glusterd_get_brick_mount_details (glusterd_brickinfo_t *brickinfo)
+glusterd_get_brick_mount_details (char *brick_path)
{
int ret = -1;
char *mnt_pt = NULL;
@@ -6994,12 +7028,12 @@ glusterd_get_brick_mount_details (glusterd_brickinfo_t *brickinfo)
this = THIS;
GF_ASSERT (this);
- GF_ASSERT (brickinfo);
+ GF_ASSERT (brick_path);
- ret = glusterd_get_brick_root (brickinfo->path, &mnt_pt);
+ ret = glusterd_get_brick_root (brick_path, &mnt_pt);
if (ret) {
gf_log (this->name, GF_LOG_ERROR, "Failed to get mount point "
- "for %s brick", brickinfo->path);
+ "for %s brick", brick_path);
goto out;
}