[
https://issues.apache.org/jira/browse/HBASE-17851?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16391362#comment-16391362
]
Josh Elser commented on HBASE-17851:
------------------------------------
Nice test overall, Vlad!
{code:java}
+ edit.add(new KeyValue(rowName, family, Bytes.toBytes("1"),
+ System.currentTimeMillis(), value));
+ long txid = log.append(info, getWalKeyImpl(System.currentTimeMillis(),
scopes), edit, true);
+ log.sync(txid);
+
+ Thread.sleep(10); // make sure 2nd edit gets a later timestamp
+
+ edit = new WALEdit();
+ edit.add(new KeyValue(rowName, family, Bytes.toBytes("2"),
+ System.currentTimeMillis(), value));
+ txid = log.append(info, getWalKeyImpl(System.currentTimeMillis(), scopes),
edit, true);{code}
Is using a "real" timestamp actually important here? Or could you just replace
the {{System.currentTimeMillis()}} calls to be {{1}}, {{2}}, and {{3}} (or
similar)?
{code:java}
+ //walfactory.shutdown(){code}
Nit: remove commented out code, please.
Otherwise, some trivial checkstyle and whitespace fixes. I think a v2 would be
ready to go!
> [TEST]: WAL to HFile conversion phase MUST detect and handle missing WAL files
> ------------------------------------------------------------------------------
>
> Key: HBASE-17851
> URL: https://issues.apache.org/jira/browse/HBASE-17851
> Project: HBase
> Issue Type: Sub-task
> Reporter: Vladimir Rodionov
> Assignee: Vladimir Rodionov
> Priority: Major
> Attachments: HBASE-17851-v1.patch
>
>
> The code is implemented - we need UT to verify correctness of a failure
> handling algo.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
