Allow a subset of failed devices to be replaced in RAID LVs.

If two devices in an array failed, it was previously impossible to replace
just one of them.  This patch allows for the replacement of some, but perhaps
not all, failed devices.

WHATS_NEW

@@ -1,5 +1,6 @@
 Version 2.02.96 - 
 ================================
+  Allow subset of failed devices to be replaced in RAID LVs.
   Prevent resume from creating error devices that already exist from suspend.
   Improve clmvd singlenode locking for better testing.
   Update and correct lvs man page with supported column names.

lib/metadata/raid_manip.c

@@ -1632,10 +1632,28 @@ int lv_raid_replace(struct logical_volume *lv,
 	 *
 	 * - We need to change the LV names when we insert them.
 	 */
+try_again:
 	if (!_alloc_image_components(lv, allocate_pvs, match_count,
 				     &new_meta_lvs, &new_data_lvs)) {
 		log_error("Failed to allocate replacement images for %s/%s",
 			  lv->vg->name, lv->name);
+
+		/*
+		 * If this is a repair, then try to
+		 * do better than all-or-nothing
+		 */
+		if (match_count > 1) {
+			log_error("Attempting replacement of %u devices"
+				  " instead of %u", match_count - 1, match_count);
+			match_count--;
+
+			/*
+			 * Since we are replacing some but not all of the bad
+			 * devices, we must set partial_activation
+			 */
+			lv->vg->cmd->partial_activation = 1;
+			goto try_again;
+		}
 		return 0;
 	}