Temptative patch to troubleshoot container id collision issue (updated at 15:15 UTC to add a mutex)

utils-use-native-urandom.diff

diff --git a/utils/utils.go b/utils/utils.go
index da6854b..0eb283a 100644
--- a/utils/utils.go
+++ b/utils/utils.go
@@ -2,12 +2,12 @@ package utils
 
 import (
 	"bytes"
-	"crypto/rand"
 	"crypto/sha1"
 	"crypto/sha256"
 	"encoding/hex"
 	"fmt"
 	"io"
+	"bufio"
 	"io/ioutil"
 	"net/http"
 	"os"
@@ -244,7 +244,7 @@ func GetTotalUsedFds() int {
 // TruncateID returns a shorthand version of a string identifier for convenience.
 // A collision with other shorthands is very unlikely, but possible.
 // In case of a collision a lookup with TruncIndex.Get() will fail, and the caller
-// will need to use a langer prefix, or the full-length Id.
+// will need to use a larger prefix, or the full-length Id.
 func TruncateID(id string) string {
 	shortLen := 12
 	if len(id) < shortLen {
@@ -253,16 +253,37 @@ func TruncateID(id string) string {
 	return id[:shortLen]
 }
 
+var urandomMutex   sync.Mutex
+
+func ReadRandomData(amount int) ([]byte, error) {
+	urandomMutex.Lock()
+	defer urandomMutex.Unlock()
+	file, err := os.Open("/dev/urandom")
+	if err != nil {
+		return nil, err
+	}
+	defer file.Close()
+
+	bytes := make([]byte, amount)
+	bufr := bufio.NewReader(file)
+	_,err = bufr.Read(bytes)
+	if err != nil {
+		return nil, err
+	}
+
+	return bytes, nil
+}
+
 // GenerateRandomID returns an unique id
 func GenerateRandomID() string {
 	for {
-		id := make([]byte, 32)
-		if _, err := io.ReadFull(rand.Reader, id); err != nil {
+		id, err := ReadRandomData(32);
+		if err != nil {
 			panic(err) // This shouldn't happen
 		}
 		value := hex.EncodeToString(id)
 		// if we try to parse the truncated for as an int and we don't have
-		// an error then the value is all numberic and causes issues when
+		// an error then the value is all numeric and causes issues when
 		// used as a hostname. ref #3869
 		if _, err := strconv.ParseInt(TruncateID(value), 10, 64); err == nil {
 			continue

Could this just be a multi-thread issue and you need to take a thread lock?