[ https://issues.apache.org/jira/browse/HBASE-17095?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15666736#comment-15666736 ]
Jingcheng Du commented on HBASE-17095: -------------------------------------- bq. We may have to do some general way? Than throw back FNFE and CorruptedHFile exception, an exception which will cause the op fully fail? Yes, we can create a new exception to wrap the CorruptHFileException and FileNotFoundException. I think it is okay to fail the operations if the file is corrupt or not found. What is your idea? bq. May be the down layer which actually see CorruptedHFile exception itself has to make the new Exception type , wrapping the original one? FNFE is in case of MOB where the ref cell resolve to some path but u r not able to find it. Correct? Correct, I found this when I debugged the MOB snapshot issues. Let me prepare another patch to address the comments. Thanks. > The ClientSimpleScanner keeps retrying if the hfile is corrupt or cannot found > ------------------------------------------------------------------------------ > > Key: HBASE-17095 > URL: https://issues.apache.org/jira/browse/HBASE-17095 > Project: HBase > Issue Type: Bug > Components: regionserver, scan > Reporter: Jingcheng Du > Assignee: Jingcheng Du > Attachments: HBASE-17095.patch, TestScannerWithCorruptHFile.java > > > In {{RsRPCServices.scan}}, most of IOE are thrown as a > {{ScannerResetException}}, even when the hfile is corrupt or it cannot be > found. The {{ClientScannr.loadCache}} will keep retrying when the exception > is {{ScannerResetException}}. We could throw CorruptHFileException and > FileNotFoundException directly from server and don't retry the scan in the > client. -- This message was sent by Atlassian JIRA (v6.3.4#6332)