[
https://issues.apache.org/jira/browse/CASSANDRA-14638?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16581293#comment-16581293
]
Aleksey Yeschenko commented on CASSANDRA-14638:
-----------------------------------------------
Unfortunately this is an accidental, lower level bug in 3.0, that I believe
should be fixed in the next minor.
The internal contract for wildcard order in 2.1, and, really, everywhere since
introduction of static columns is the following:
1. Partition key columns in their positional order
2. Clustering columns in their positional order
3. All static columns, ordered alphabetically by name, irregardless of whether
they are simple or complex
4. All regular columns, ordered alphabetically by name, irregardless of whether
they are simple or complex
To illustrate, let's say we have a table defined as
{code}
CREATE TABLE ks.tbl (p0 text, p1 text, a set<text>, b set<text> static, c text,
d text static, e set<text>, f set<text> static, g text, h text static, i
set<text>, j set<text> static, PRIMARY KEY (p0, p1));
{code}
And one inserted row:
{code}
INSERT INTO ks.tbl (p0, p1, a, b, c, d, e, f, g, h, i, j) VALUES ('p0', 'p1',
{'a'}, {'b'}, 'c', 'd', {'e'}, {'f'}, 'g', 'h', {'i'}, {'j'});
{code}
In 2.1, performing a {{SELECT * FROM ks.tbl;}} query would return the following
result set:
{code}
p0 | p1 | b | d | f | h | j | a | c | e | g | i
----+----+-------+---+-------+---+-------+-------+---+-------+---+-------
p0 | p1 | {'b'} | d | {'f'} | h | {'j'} | {'a'} | c | {'e'} | g | {'i'}
{code}
3.0, in contrast, would return the following result set:
{code}
p0 | p1 | d | h | b | f | j | a | c | e | g | i
----+----+---+---+-------+-------+-------+-------+---+-------+---+-------
p0 | p1 | d | h | {'b'} | {'f'} | {'j'} | {'a'} | c | {'e'} | g | {'i'}
{code}
Both correctly return {{PRIMARY KEY}} columns in their defined order first,
then all the static columns (although 3.0 order for them is all wrong),
followed by all regular columns in alphabetic order, correct in both 2.1 and
3.0.
What 3.0 does incorrectly is ordering the static columns it returns. Instead of
merging simple and complex static columns, we are getting all simple static
columns first, followed by all complex static columns.
The bug is deep in {{Columns}} class, more specifically in
{{findFirstComplexIdx()}} method. To find the first complex column index we are
using a surrogate {{ColumnDefinition}} value {{FIRST_COMPLEX}}, which is of
kind {{REGULAR}} for {{Columns}} of both {{REGULAR}} and {{STATIC}} kinds. This
quite clearly breaks {{Columns}} of kind {{STATIC}}, making the container
believe that all of its columns are {{COMPLEX}}.
> Column result order can change in 'SELECT *' results when upgrading from 2.1
> to 3.0 causing response corruption for queries using prepared statements when
> static columns are used
> ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: CASSANDRA-14638
> URL: https://issues.apache.org/jira/browse/CASSANDRA-14638
> Project: Cassandra
> Issue Type: Bug
> Environment: Single C* node ccm cluster upgraded from C* 2.1.20 to
> 3.0.17
> Reporter: Andy Tolbert
> Assignee: Aleksey Yeschenko
> Priority: Major
>
> When performing an upgrade from C* 2.1.20 to 3.0.17 I observed that the order
> of columns returned from a 'SELECT *' query changes, particularly when static
> columns are involved.
> This may not seem like that much of a problem, however if using Prepared
> Statements, any clients that remain connected during the upgrade may
> encounter issues consuming results from these queries, as data is reordered
> and the client not aware of it. The result definition is sent in the
> original prepared statement response, so if order changes the client has no
> way of knowing (until C* 4.0 via CASSANDRA-10786) without re-preparing, which
> is non-trivial as most client drivers cache prepared statements.
> This could lead to reading the wrong values for columns, which could result
> in some kind of deserialization exception or if the data types of the
> switched columns are compatible, the wrong values. This happens even if the
> client attempts to retrieve a column value by name (i.e. row.getInt("colx")).
> Unfortunately I don't think there is an easy fix for this. If the order was
> changed back to the previous format, you risk issues for users upgrading from
> older 3.0 version. I think it would be nice to add a note in the NEWS file
> in the 3.0 upgrade section that describes this issue, and how to work around
> it (specify all column names of interest explicitly in query).
> Example schema and code to reproduce:
>
> {noformat}
> create keyspace ks with replication = {'class': 'SimpleStrategy',
> 'replication_factor': 1};
> create table ks.tbl (p0 text,
> p1 text,
> m map<text, text> static,
> t text,
> u text static,
> primary key (p0, p1)
> );
> insert into ks.tbl (p0, p1, m, t, u) values ('p0', 'p1', { 'm0' : 'm1' },
> 't', 'u');{noformat}
>
> When querying with 2.1 you'll observe the following order via cqlsh:
> {noformat}
> p0 | p1 | m | u | t
> ----+----+--------------+---+---
> p0 | p1 | {'m0': 'm1'} | u | t{noformat}
>
> With 3.0, observe that u and m are transposed:
>
> {noformat}
> p0 | p1 | u | m | t
> ----+----+---+--------------+---
> p0 | p1 | u | {'m0': 'm1'} | t{noformat}
>
>
> {code:java}
> import com.datastax.driver.core.BoundStatement;
> import com.datastax.driver.core.Cluster;
> import com.datastax.driver.core.ColumnDefinitions;
> import com.datastax.driver.core.PreparedStatement;
> import com.datastax.driver.core.ResultSet;
> import com.datastax.driver.core.Row;
> import com.datastax.driver.core.Session;
> import com.google.common.util.concurrent.Uninterruptibles;
> import java.util.concurrent.TimeUnit;
> public class LiveUpgradeTest {
> public static void main(String args[]) {
> Cluster cluster = Cluster.builder().addContactPoints("127.0.0.1").build();
> try {
> Session session = cluster.connect();
> PreparedStatement p = session.prepare("SELECT * from ks.tbl");
> BoundStatement bs = p.bind();
> // continually query every 30 seconds
> while (true) {
> try {
> ResultSet r = session.execute(bs);
> Row row = r.one();
> int i = 0;
> // iterate over the result metadata in order printing the
> // index, name, type, and length of the first row of data.
> for (ColumnDefinitions.Definition d : r.getColumnDefinitions()) {
> System.out.println(
> i++
> + ": "
> + d.getName()
> + " -> "
> + d.getType()
> + " -> val = "
> + row.getBytesUnsafe(d.getName()).array().length);
> }
> } catch (Throwable t) {
> t.printStackTrace();
> } finally {
> Uninterruptibles.sleepUninterruptibly(30, TimeUnit.SECONDS);
> }
> }
> } finally {
> cluster.close();
> }
> }
> }
> {code}
> To reproduce, set up a cluster, the schema, and run this script. Then
> upgrade the cluster to 3.0.17 (with ccm, ccm stop; ccm node1 setdir -v
> 3.0.17; ccm start works) and observe after the client is able to reconnect
> that the results are in a different order. i.e.:
>
> With 2.x:
>
> {noformat}
> 0: p0 -> varchar -> val = 2
> 1: p1 -> varchar -> val = 2
> 2: m -> map<varchar, varchar> -> val = 16
> 3: u -> varchar -> val = 1
> 4: t -> varchar -> val = 1{noformat}
>
> With 3.x:
>
> {noformat}
> 0: p0 -> varchar -> val = 2
> 1: p1 -> varchar -> val = 2
> 2: m -> map<varchar, varchar> -> val = 1
> 3: u -> varchar -> val = 16 (<-- the data for 'm' is now at index 3)
> 4: t -> varchar -> val = 1{noformat}
>
>
>
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
