[
https://issues.apache.org/jira/browse/PHOENIX-4156?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16155622#comment-16155622
]
Samarth Jain commented on PHOENIX-4156:
---------------------------------------
[~jamestaylor] - looks like the test is still flapping a bit but only for one
case:
{code}
testIndexWriteFailure[MutableIndexFailureIT_transactional=false,localIndex=false,isNamespaceMapped=false,disableIndexOnWriteFailure=false,rebuildIndexOnWriteFailure=false,failRebuildTask=false,throwIndexWriteFailure=null]
{code}
It looks like in the case when rebuildIndexOnWriteFailure=false, we are
replaying mutations by issuing DMLs with SCN:
{code}
private void replayMutations() throws SQLException {
Properties props = PropertiesUtil.deepCopy(TEST_PROPERTIES);
for (int i = 0; i < exceptions.size(); i++) {
CommitException e = exceptions.get(i);
long ts = e.getServerTimestamp();
props.setProperty(PhoenixRuntime.REPLAY_AT_ATTRIB,
Long.toString(ts));
try (Connection conn = DriverManager.getConnection(getUrl(),
props)) {
if (i == 0) {
updateTable(conn, false);
} else if (i == 1) {
updateTableAgain(conn, false);
} else {
fail();
}
}
}
}
{code}
Considering we are working on disallowing DMLs with SCN, it makes this test
case invalid. I guess it also makes the mode rebuildIndexOnWriteFailure=false
also not as useful because now clients won't be able to reply the mutations to
get the index back in sync. Should we remove this case altogether from this
test?
> Fix flapping MutableIndexFailureIT
> ----------------------------------
>
> Key: PHOENIX-4156
> URL: https://issues.apache.org/jira/browse/PHOENIX-4156
> Project: Phoenix
> Issue Type: Bug
> Reporter: Samarth Jain
> Assignee: Samarth Jain
> Attachments: PHOENIX-4156_v1.patch, PHOENIX-4156_v2.patch,
> PHOENIX-4156_v3.patch, PHOENIX-4156_v4.patch, PHOENIX-4156_v5.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
