Geoffrey Jacoby created PHOENIX-5541:
----------------------------------------
Summary: Redundant Global Mutable Index Writes
Key: PHOENIX-5541
URL: https://issues.apache.org/jira/browse/PHOENIX-5541
Project: Phoenix
Issue Type: Bug
Affects Versions: 4.14.3, 4.15.0
Reporter: Geoffrey Jacoby
Assignee: Geoffrey Jacoby
In the inaccurately-named IndexWriter.writeAndKillYourselfOnFailure, there's
the following code:
{code:java}
public void writeAndKillYourselfOnFailure(Collection<Pair<Mutation, byte[]>>
indexUpdates, boolean allowLocalUpdates, int clientVersion) throws IOException
{
// convert the strings to htableinterfaces to which we can talk and group by
TABLE Multimap<HTableInterfaceReference, Mutation> toWrite =
resolveTableReferences(indexUpdates);
writeAndKillYourselfOnFailure(toWrite, allowLocalUpdates, clientVersion);
writeAndHandleFailure(toWrite, allowLocalUpdates, clientVersion);
}
{code}
writeAndKillYourselfOnFailure and writeAndHandleFailure seem to be identical,
which means that calling them both will result in the same index Cells being
written twice. This shouldn't affect correctness, but it will affect
performance and (temporarily, until compaction) storage
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
