Details
-
Bug
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
2.2.3
-
None
Description
While testing with large data, it is observed that some time hbck page was not accessible and NPE was thrown in log.
【Test step】Excat Steps are not known. Observed this while executing below steps
1: Test data: 9K regions,~300 issues in my test setup
2: run hbck_chore and catalogjanitor_run multiple times
3: Also run balancer so that region movement is happening.
4: Observe that HBCK page not available
2020-09-21 20:05:31,706 WARN [qtp1698322791-94] server.HttpChannel: /hbck.jsp
java.lang.NullPointerException
at org.apache.hadoop.hbase.generated.master.hbck_jsp.formatServerName(hbck_jsp.java:37)
at org.apache.hadoop.hbase.generated.master.hbck_jsp._jspService(hbck_jsp.java:144)
at org.apache.jasper.runtime.HttpJspBase.service(HttpJspBase.java:111)
at javax.servlet.http.HttpServlet.service(HttpServlet.java:790)
at org.eclipse.jetty.servlet.ServletHolder.handle(ServletHolder.java:844)
Attachments
Issue Links
- links to
Seems problem is originating from HbckChore.loadRegionsFromRSReport()
While populating inconsistent regions, locationInMeta null check is done only for the case where hri.getDeployedOn().size() == 0 for other cases locationInMeta null check is not done.
To fix the issue, we should do the null check for all the inconsistent region cases.