[
https://issues.apache.org/jira/browse/HIVE-23671?focusedWorklogId=455371&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-455371
]
ASF GitHub Bot logged work on HIVE-23671:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 07/Jul/20 11:26
Start Date: 07/Jul/20 11:26
Worklog Time Spent: 10m
Work Description: pvargacl commented on a change in pull request #1087:
URL: https://github.com/apache/hive/pull/1087#discussion_r450793830
##########
File path:
standalone-metastore/metastore-server/src/main/java/org/apache/hadoop/hive/metastore/txn/TxnHandler.java
##########
@@ -2032,28 +2077,61 @@ public void
seedWriteIdOnAcidConversion(InitializeTableWriteIdsRequest rqst)
// The initial value for write id should be 1 and hence we add 1 with
number of write ids
// allocated here
String s = "INSERT INTO \"NEXT_WRITE_ID\" (\"NWI_DATABASE\",
\"NWI_TABLE\", \"NWI_NEXT\") VALUES (?, ?, "
- + Long.toString(rqst.getSeeWriteId() + 1) + ")";
- pst = sqlGenerator.prepareStmtWithParameters(dbConn, s,
Arrays.asList(rqst.getDbName(), rqst.getTblName()));
+ + Long.toString(rqst.getSeedWriteId() + 1) + ")";
+ pst = sqlGenerator.prepareStmtWithParameters(dbConn, s,
Arrays.asList(rqst.getDbName(), rqst.getTableName()));
LOG.debug("Going to execute insert <" + s.replaceAll("\\?", "{}") +
">",
- quoteString(rqst.getDbName()), quoteString(rqst.getTblName()));
+ quoteString(rqst.getDbName()),
quoteString(rqst.getTableName()));
pst.execute();
LOG.debug("Going to commit");
dbConn.commit();
} catch (SQLException e) {
- LOG.debug("Going to rollback");
rollbackDBConn(dbConn);
- checkRetryable(dbConn, e, "seedWriteIdOnAcidConversion(" + rqst + ")");
- throw new MetaException("Unable to update transaction database "
- + StringUtils.stringifyException(e));
+ checkRetryable(dbConn, e, "seedWriteId(" + rqst + ")");
+ throw new MetaException("Unable to update transaction database " +
StringUtils.stringifyException(e));
} finally {
close(null, pst, dbConn);
unlockInternal();
}
} catch (RetryException e) {
- seedWriteIdOnAcidConversion(rqst);
+ seedWriteId(rqst);
}
+ }
+
+ @Override
+ public void seedTxnId(SeedTxnIdRequest rqst) throws MetaException {
+ try {
+ Connection dbConn = null;
+ Statement stmt = null;
+ try {
+ lockInternal();
+ dbConn = getDbConn(Connection.TRANSACTION_READ_COMMITTED);
+ stmt = dbConn.createStatement();
+ /*
+ * Locking the txnLock an exclusive way, we do not want to set the
txnId backward accidentally
+ * if there are concurrent open transactions
+ */
+ acquireTxnLock(stmt, false);
+ long highWaterMark = getHighWaterMark(stmt);
+ if (highWaterMark >= rqst.getSeedTxnId()) {
Review comment:
This is not about the writeIds, it is about the txnId. If you have a
data from a database where there were high amount of transaction and the
compaction run on the table, you will have some high txnId in the
visibilityTxnId in the name of the compacted folder.
If you then move this data to a cluster with less transaction (ex. a test
cluster) and you run the msck repair, you have to skip the txnId forward so the
next query will read the compacted folder. Here the race condition is, that
somehow the txnId sequence gets ahead of you between the check and the seeding
the value, in that case we throw this exception to not to set the sequence
backward. Anyway, in this case if you run the msck repair again it will
succeed, since the txnid will be high enough.
The writeId race condition won't cause a problem I think, since if some
other transaction allocated the first writeId the seedWriteId will fail on the
unique constraint on the table
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 455371)
Time Spent: 11.5h (was: 11h 20m)
> MSCK repair should handle transactional tables in certain usecases
> ------------------------------------------------------------------
>
> Key: HIVE-23671
> URL: https://issues.apache.org/jira/browse/HIVE-23671
> Project: Hive
> Issue Type: Improvement
> Components: Metastore
> Reporter: Peter Varga
> Assignee: Peter Varga
> Priority: Major
> Labels: pull-request-available
> Time Spent: 11.5h
> Remaining Estimate: 0h
>
> The MSCK REPAIR tool does not handle transactional tables too well. It can
> find and add new partitions the same way as for non-transactional tables, but
> since the writeId differences are not handled, the data can not read back
> from the new partitions.
> We could handle some usecases when the writeIds in the HMS and the underlying
> data are not conflicting. If the HMS does not contains allocated writes for
> the table we can seed the table with the writeIds read from the directory
> structrure.
> Real life use cases could be:
> * Copy data files from one cluster to another with different HMS, create the
> table and call MSCK REPAIR
> * If the HMS db is lost, recreate the table and call MSCK REPAIR
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
