Jody Landreneau created PHOENIX-1181:
----------------------------------------
Summary: client cache fails to update itself after a table was
altered from a diff client
Key: PHOENIX-1181
URL: https://issues.apache.org/jira/browse/PHOENIX-1181
Project: Phoenix
Issue Type: Bug
Affects Versions: 3.0.0, 3.1
Environment: I have tried on both the 3.0 and 3.1 latest as of
2014/08/18
Reporter: Jody Landreneau
Description:
If you have multiple phoenix clients running, which could be on several
physical machines(diff vms), changing the schema of a table(by adding or
removing a field) will result in errors in the clients that didn't issue the
alter. This appears to be due to an internal client cache that is not
refreshed. I note that the connections get their cache from this shared client
cache so creating/closing connections does not help.
Repro:
1) A somewhat simple way to repro is to open the SQuirrel client and issue the
following:
create table if not exists test_simpletable (
id VARCHAR NOT NULL,
field1 BIGINT
CONSTRAINT pk PRIMARY KEY (id))
2) use the following code snippet to run a client that inserts into the simple
table and then pauses, allowing you to issue an alter in the SQuirrel client by
adding an additional field("alter table test_simpletable add field2 BIGINT").
It then proceeds to attempt another insert.
{code}
public class ConnCacheTest {
public static final String PHOENIX_JDBC_URL = "jdbc:phoenix:localhost";
public static final String INSERT_ONE_FIELD = "upsert into test_simpletable
(id, field1) values ( ?, ?)";
public static final String INSERT_TWO_FIELD = "upsert into test_simpletable
(id, field1, field2) values ( ?, ?, ?)";
public static void main(String[] args){
PhoenixConnection conn = null;
try {
Class.forName("org.apache.phoenix.jdbc.PhoenixDriver");
Properties props = new Properties();
// here we insert into the orig schema with one column
conn = (PhoenixConnection)
DriverManager.getConnection(PHOENIX_JDBC_URL, props);
PreparedStatement stmtInsert1 =
conn.prepareStatement(INSERT_ONE_FIELD);
stmtInsert1.setString(1, "key1");
stmtInsert1.setLong(2, 1L);
stmtInsert1.execute();
conn.commit();
stmtInsert1.close();
conn.close();
// While sleeping for a min, alter the table to have an additional
// column. Do the alter through a separate client like SQuirreL.
System.out.println("Starting to wait, make the alter!!!");
Thread.sleep(1000*60*1);
// this insert will try to insert to two columns
conn = (PhoenixConnection)
DriverManager.getConnection(PHOENIX_JDBC_URL, props);
PreparedStatement pstmt2 = conn.prepareStatement(INSERT_TWO_FIELD);
pstmt2.setString(1, "key2");
pstmt2.setLong(2, 2L);
pstmt2.setLong(3, 2L);
pstmt2.execute();
conn.commit();
pstmt2.close();
conn.close();
} catch (Exception ex) {
ex.printStackTrace();
}
}
}
{code}
Result:
org.apache.phoenix.schema.ColumnNotFoundException: ERROR 504 (42703): Undefined
column. columnName=FIELD2
at org.apache.phoenix.schema.PTableImpl.getColumn(PTableImpl.java:513)
at
org.apache.phoenix.compile.FromCompiler$SingleTableColumnResolver.resolveColumn(FromCompiler.java:254)
at
org.apache.phoenix.compile.UpsertCompiler.compile(UpsertCompiler.java:307)
at
org.apache.phoenix.jdbc.PhoenixStatement$ExecutableUpsertStatement.compilePlan(PhoenixStatement.java:442)
at
org.apache.phoenix.jdbc.PhoenixStatement$ExecutableUpsertStatement.compilePlan(PhoenixStatement.java:433)
at
org.apache.phoenix.jdbc.PhoenixStatement$2.call(PhoenixStatement.java:250)
at
org.apache.phoenix.jdbc.PhoenixStatement$2.call(PhoenixStatement.java:242)
at
org.apache.phoenix.util.PhoenixContextExecutor.call(PhoenixContextExecutor.java:54)
at
org.apache.phoenix.jdbc.PhoenixStatement.executeMutation(PhoenixStatement.java:241)
at
org.apache.phoenix.jdbc.PhoenixStatement.execute(PhoenixStatement.java:190)
at
org.apache.phoenix.jdbc.PhoenixPreparedStatement.execute(PhoenixPreparedStatement.java:147)
at
org.apache.phoenix.jdbc.PhoenixPreparedStatement.execute(PhoenixPreparedStatement.java:152)
Expected:
I would expect there to be some timeout setting on the client cache(I believe
the server side has this setting) or when a new connection is created it gets
the latest schema from metadata? This would allow users to set time on their
connections.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
