Github user selvaganesang commented on a diff in the pull request:
https://github.com/apache/incubator-trafodion/pull/730#discussion_r81178408
--- Diff:
core/sqf/src/seatrans/tm/hbasetmlib2/src/main/java/org/trafodion/dtm/HBaseTxClient.java
---
@@ -1034,34 +1034,23 @@ public void run() {
}
try {
TxRecoverList =
txnManager.recoveryRequest(hostnamePort, regionBytes, tmID);
- }catch (NotServingRegionException e) {
- TxRecoverList = null;
- LOG.error("TRAF RCOV
THREAD:NotServingRegionException calling recoveryRequest. regionBytes: " + new
String(regionBytes) +
- " TM: " + tmID + "
hostnamePort: " + hostnamePort, e);
-
- // First delete the zookeeper entry
- LOG.error("TRAF RCOV
THREAD:recoveryRequest. Deleting region entry Entry: " + regionEntry);
-
zookeeper.deleteRegionEntry(regionEntry);
- // Create a local HTable object
using the regionInfo
- HTable table = new HTable(config,
HRegionInfo.parseFrom(regionBytes).getTable().getNameAsString());
- // Repost a zookeeper entry for all
current regions in the table
-
zookeeper.postAllRegionEntries(table);
- }// NotServingRegionException
- catch (TableNotFoundException tnfe) {
- // In this case there is nothing to
recover. We just need to delete the region entry.
- // First delete the zookeeper entry
- LOG.warn("TRAF RCOV
THREAD:TableNotFoundException calling txnManager.recoveryRequest. " + "TM: " +
- tmID + " regionBytes: [" +
regionBytes + "]. Deleting zookeeper region entry. \n exception: " + tnfe);
-
zookeeper.deleteRegionEntry(regionEntry);
-
- }// TableNotFoundException
- catch (DeserializationException de) {
- // We are unable to parse the region
info from ZooKeeper We just need to delete the region entry.
- // First delete the zookeeper entry
- LOG.warn("TRAF RCOV
THREAD:DeserializationException calling txnManager.recoveryRequest. " + "TM: " +
- tmID + " regionBytes: [" +
regionBytes + "]. Deleting zookeeper region entry. \n exception: " + de);
-
zookeeper.deleteRegionEntry(regionEntry);
- }// DeserializationException
+ }
+ catch (IOException e) {
+ // For all cases of Exception, we rely
on the region to redrive the request.
+ // Likely there is nothing to recover,
due to a stale region entry, but it is always safe to redrive.
+ // We log a warning event and delete
the ZKNode entry.
+ LOG.warn("TRAF RCOV THREAD:Exception
calling txnManager.recoveryRequest. " + "TM: " +
+ tmID + " regionBytes: [" +
regionBytes + "]. Deleting zookeeper region entry. \n exception: ", e);
+
zookeeper.deleteRegionEntry(regionEntry);
+
+ // In the case of
NotServingRegionException we will repost the ZKNode after refreshing the table.
+ if (e instanceof
NotServingRegionException){
--- End diff --
Do you think we should consider checking e.getCause() instanceof
NotServingRegionException too. Rationale: Assuming recoveryRequest throws
NotServingRegionException, it would be caught in 3018 and transformed into
IOException with cause as NotServingRegionException.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
