CassandraCAS

Compare-and-swap in Cassandra

Setup

From Cassandra cqlsh:

CREATE KEYSPACE IF NOT EXISTS datomic WITH replication = {'class': 'SimpleStrategy', 'replication_factor' : 3};
CREATE TABLE IF NOT EXISTS datomic.datomic
(
  id text PRIMARY KEY,
  rev bigint,
  map text,
  val blob
);

Start a 3+ node cassandra cluster. Get Cassandra running locally on 127.0.0.1, 2, 3. CCM is great for this.

Building

mvn clean package

Running CAS Stress Test

The program

java -jar target/CassandraCAS-0.0.jar {host} {port} raceSessions {nthreads} {msec} {target}

will start nthreads racing to CAS a counter from 0 up to target. Each thread will pause for msec between operations (or pass 0 msec for no pause).

Running Multiple Threads on One Client Process

The following example is enough load to force some write retries on my local dev box:

java -jar target/CassandraCAS-0.0.jar 127.0.0.1 9042 create
java -jar target/CassandraCAS-0.0.jar 127.0.0.1 9042 raceSessions 5 0 500

Running Multiple Client Processes

Setup

java -jar target/CassandraCAS-0.0.jar 127.0.0.1 9042 create

Once per client (need to start at about same time or first one can finish before others begin)

java -jar target/CassandraCAS-0.0.jar 127.0.0.1 9042 raceSessions 1 0 5000

Validating CAS

• Collect the outputs from one or more clients running CAS stress test into a single file.
• Load the com.datomic.validate-cas namespace in a Clojure REPL. No deps, so vanilla Clojure REPL will do.
• Run validate-file against the output file. If the :valid? key is false, the validator thinks a consistency violation has occurred.

Java Retry Bug

• When a write timeout occurs, the Java driver can be configured to call a RetryPolicy to allow clients flexibility in error handling.
• The onWriteTimeout method is passed a single ConsistencyLevel argument, described as "the original consistency level of the write that timed out." This looks a bit sketchy since writes can set two kinds of consistency levels (CL and SCL), but the retry handler only gets informed about one.
• When a write timeout occurs on a SCL=SERIAL write, the consistency level passed to onWriteTimeout is SERIAL. Retrying with the level given results in an InvalidQueryException.

In short, a query that is valid on initial try is deemed invalid on retry. You can usually see this by running with the following settings

java -Dcom.datomic.CassandraCASRetryPolicy=Documented -jar target/CassandraCAS-0.0.jar 127.0.0.1 9042 raceSessions 5 0 500

Thread Safety Testing

Setup

mvn clean package
java -jar target/CassandraCAS-0.0.jar 127.0.0.1 9042 create

First use of driver is off main thread: not ok

java -jar target/CassandraCAS-0.0.jar 127.0.0.1 9042 casThread

com.datastax.driver.core.exceptions.NoHostAvailableException: All host(s) tried for query failed (no host was tried)
    at com.datastax.driver.core.exceptions.NoHostAvailableException.copy(NoHostAvailableException.java:84)
    at com.datastax.driver.core.DefaultResultSetFuture.extractCauseFromExecutionException(DefaultResultSetFuture.java:289)
    at com.datastax.driver.core.DefaultResultSetFuture.getUninterruptibly(DefaultResultSetFuture.java:205)
    at com.datastax.driver.core.AbstractSession.execute(AbstractSession.java:52)
    at com.datomic.CassandraCAS.casRev(CassandraCAS.java:142)
    at com.datomic.CassandraCAS$5.run(CassandraCAS.java:263)
    at java.lang.Thread.run(Thread.java:745)
Caused by: com.datastax.driver.core.exceptions.NoHostAvailableException: All host(s) tried for query failed (no host was tried)
    at com.datastax.driver.core.RequestHandler.sendRequest(RequestHandler.java:107)
    at com.datastax.driver.core.SessionManager.execute(SessionManager.java:538)
    at com.datastax.driver.core.SessionManager.executeQuery(SessionManager.java:577)
    at com.datastax.driver.core.SessionManager.executeAsync(SessionManager.java:119)

First use of driver is on main thread, then CAS from another thread: ok

java -jar target/CassandraCAS-0.0.jar 127.0.0.1 9042 readMainThenCASThread

CAS to 1

First use of driver is on main thread, then another thread touches mbean then does CAS: not ok

java -jar target/CassandraCAS-0.0.jar 127.0.0.1 9042 readMainThenMbeanCASThread

CAS to 2
com.datastax.driver.core.exceptions.NoHostAvailableException: All host(s) tried for query failed (no host was tried)
    at com.datastax.driver.core.exceptions.NoHostAvailableException.copy(NoHostAvailableException.java:84)
    at com.datastax.driver.core.DefaultResultSetFuture.extractCauseFromExecutionException(DefaultResultSetFuture.java:289)
    at com.datastax.driver.core.DefaultResultSetFuture.getUninterruptibly(DefaultResultSetFuture.java:205)
    at com.datastax.driver.core.AbstractSession.execute(AbstractSession.java:52)
    at com.datomic.CassandraCAS.casRev(CassandraCAS.java:142)
    at com.datomic.CassandraCAS$6.run(CassandraCAS.java:278)
    at java.lang.Thread.run(Thread.java:745)
Caused by: com.datastax.driver.core.exceptions.NoHostAvailableException: All host(s) tried for query failed (no host was tried)
    at com.datastax.driver.core.RequestHandler.sendRequest(RequestHandler.java:107)
    at com.datastax.driver.core.SessionManager.execute(SessionManager.java:538)
    at com.datastax.driver.core.SessionManager.executeQuery(SessionManager.java:577)
    at com.datastax.driver.core.SessionManager.executeAsync(SessionManager.java:119)
    ... 4 more

Copyright

Copyright (c) Cognitect, Inc. All rights reserved.