[HBASE-15632] Undo the checking of lastStoreFlushTimeMap.isEmpty() introduced in HBASE-13145

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 2.0.0
• Fix Version/s: 1.3.0, 2.0.0
• Component/s: regionserver
• Assignee: Hua Xiang
• Reporter: Hua Xiang
• Created: 11/Apr/16 18:19
• Updated: 09/Nov/17 00:39
• Resolved: 12/Apr/16 19:02

Description

HBASE-13145 introduce the following check

diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegion.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegion.java
index 215069c..8f73af5 100644
--- a/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegion.java
+++ b/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegion.java
@@ -1574,7 +1574,8 @@ public class HRegion implements HeapSize, PropagatingConfigurationObserver { //
    */
   @VisibleForTesting
   public long getEarliestFlushTimeForAllStores() {
-    return Collections.min(lastStoreFlushTimeMap.values());
+    return lastStoreFlushTimeMap.isEmpty() ? Long.MAX_VALUE : Collections.min(lastStoreFlushTimeMap
+        .values());
   }
 

I think the reason for the check is that table creation without family is allowed before HBASE-15456. With HBASE-15456, table creation without family is not allowed. We have one user claimed that they run into the same HRegionServer$PeriodicMemstoreFlusher exception, and the table was created with family. The log was not kept so could not find more info there. By checking the code, it seems impossible. Can we undo this check so the real issue is not hidden in case there is one, Apache9?