Initial learning into apache cassandra paxos

Datetime:2016-08-23 00:41:49          Topic: Cassandra           Share

Recently I have been reading into apache lightweight transaction in cassandra 2.0 and interested into how it implemented in code level. From end user perspective, when you manupulating data either insert and update with if not exists, then internally, paxos operation willl be used.

An example of lightweight transaction.

INSERT INTO USERS (login, email, name, login_count) values ('jbellis', 'jbellis@datastax.com', 'Jonathan Ellis', 1) IF NOT EXISTS  
   
 UPDATE users SET reset_token = null, password = ‘newpassword’ WHERE login = ‘jbellis’ IF reset_token = ‘some-generated-reset-token’

Essentially the paxos how operation concerntrated in class StorageProxy .We read that from the code documentation,

There are three phases to Paxos:

1. Prepare: the coordinator generates a ballot (timeUUID in our case) and asks replicas to (a) promise

not to accept updates from older ballots and (b) tell us about the most recent update it has already

accepted.

2. Accept: if a majority of replicas reply, the coordinator asks replicas to accept the value of the

highest proposal ballot it heard about, or a new value if no in-progress proposals were reported.

3. Commit (Learn): if a majority of replicas acknowledge the accept request, we can commit the new

value.

So it involve a few operation before an insert and update can be perform and this is not something you want to replace in bulk operation call with. We see that in class  StorageService , several paxos verbs are registered. You can find them mostly in the paxos packages. Following are some useful paxos classes.

https://github.com/apache/cassandra/blob/cassandra-2.0.7/src/java/org/apache/cassandra/service/paxos/PaxosState.java

https://github.com/apache/cassandra/blob/cassandra-2.0.7/src/java/org/apache/cassandra/service/paxos/Commit.java

https://github.com/apache/cassandra/blob/cassandra-2.0.7/src/java/org/apache/cassandra/service/paxos/PrepareVerbHandler.java

https://github.com/apache/cassandra/blob/cassandra-2.0.7/src/java/org/apache/cassandra/service/paxos/ProposeVerbHandler.java

https://github.com/apache/cassandra/blob/cassandra-2.0.7/src/java/org/apache/cassandra/service/paxos/CommitVerbHandler.java

https://github.com/apache/cassandra/blob/cassandra-2.0.7/src/java/org/apache/cassandra/service/paxos/AbstractPaxosCallback.java

https://github.com/apache/cassandra/blob/cassandra-2.0.7/src/java/org/apache/cassandra/service/paxos/ProposeCallback.java

https://github.com/apache/cassandra/blob/cassandra-2.0.7/src/java/org/apache/cassandra/service/paxos/PrepareCallback.java

https://github.com/apache/cassandra/blob/cassandra-2.0.7/src/java/org/apache/cassandra/service/paxos/PrepareResponse.java

Interesting in the class PaxosState, noticed the following

locks - an array of length 1024

call system keyspace class to load and save paxos state.

When you check into cassandra using cqlsh, you will find the following.

cqlsh:jw_schema1> use system;  
 cqlsh:system> desc tables;  
   
 available_ranges     peers        paxos      range_xfers  
 batches          compaction_history batchlog    local     
 "IndexInfo"        sstable_activity  size_estimates hints     
 views_builds_in_progress peer_events     built_views    
   
 cqlsh:system> select * from paxos;  
   
  row_key | cf_id | in_progress_ballot | most_recent_commit | most_recent_commit_at | most_recent_commit_version | proposal | proposal_ballot | proposal_version  
 ---------+-------+--------------------+--------------------+-----------------------+----------------------------+----------+-----------------+------------------  
   
 (0 rows)

You can also read the unit test for paxos in cassandra 2.0.17 as can be read here .

This compare and set (cas) operation looks interesting if you want to ensure the value only be created if it not exists, a nifty feature found in apache cassandra 2.0 onward. Feel free to explore further!





About List