Improve init messages for distributed setup (#5786)

Fixes #5531

cmd/format-xl.go

@@ -314,6 +314,11 @@ func shouldInitXLDisks(errs []error) bool {
 	return countErrs(errs, errUnformattedDisk) == len(errs)
 }
 
+// Check if unformatted disks are equal to write quorum.
+func quorumUnformattedDisks(errs []error) bool {
+	return countErrs(errs, errUnformattedDisk) >= (len(errs)/2)+1
+}
+
 // loadFormatXLAll - load all format config from all input disks in parallel.
 func loadFormatXLAll(storageDisks []StorageAPI) ([]*formatXLV3, []error) {
 	// Initialize sync waitgroup.

cmd/prepare-storage.go

@@ -145,13 +145,23 @@ func connectLoadInitFormats(firstDisk bool, endpoints EndpointList, setCount, dr
 		}
 	}
 
-	if shouldInitXLDisks(sErrs) {
+	// All disks report unformatted we should initialized everyone.
-		if !firstDisk {
+	if shouldInitXLDisks(sErrs) && firstDisk {
-			return nil, errNotFirstDisk
-		}
 		return initFormatXL(context.Background(), storageDisks, setCount, drivesPerSet)
 	}
 
+	// Return error when quorum unformatted disks - indicating we are
+	// waiting for first server to be online.
+	if quorumUnformattedDisks(sErrs) && !firstDisk {
+		return nil, errNotFirstDisk
+	}
+
+	// Return error when quorum unformatted disks but waiting for rest
+	// of the servers to be online.
+	if quorumUnformattedDisks(sErrs) && firstDisk {
+		return nil, errFirstDiskWait
+	}
+
 	// Following function is added to fix a regressions which was introduced
 	// in release RELEASE.2018-03-16T22-52-12Z after migrating v1 to v2 to v3.
 	// This migration failed to capture '.This' field properly which indicates
@@ -219,6 +229,10 @@ func waitForFormatXL(ctx context.Context, firstDisk bool, endpoints EndpointList
 					// Fresh setup, wait for first server to be up.
 					logger.Info("Waiting for the first server to format the disks.")
 					continue
+				case errFirstDiskWait:
+					// Fresh setup, wait for other servers to come up.
+					logger.Info("Waiting for all other servers to be online to format the disks.")
+					continue
 				case errXLReadQuorum:
 					// no quorum available continue to wait for minimum number of servers.
 					logger.Info("Waiting for a minimum of %d disks to come online (elapsed %s)\n", len(endpoints)/2, getElapsedTime())

cmd/typed-errors.go

@@ -62,3 +62,6 @@ var errInvalidRangeSource = errors.New("Range specified exceeds source object si
 // error returned by disks which are to be initialized are waiting for the
 // first server to initialize them in distributed set to initialize them.
 var errNotFirstDisk = errors.New("Not first disk")
+
+// error returned by first disk waiting to initialize other servers.
+var errFirstDiskWait = errors.New("Waiting on other disks")