[jira] [Commented] (CASSANDRA-2686) Distributed per row locks
[ https://issues.apache.org/jira/browse/CASSANDRA-2686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13053296#comment-13053296 ] Luís Ferreira commented on CASSANDRA-2686: -- I've taken your advice and used Cages for the Read and Write locks. From this I constructed a Transaction system on top of Cassandra. As soon as I have some performance test results I'll put them here, as well as the code, if anyone is interested. It basically implements a write ahead log, taking advantage of the atomicity in per row updates, and of idempotent updates. It also has a pre processing mechanism for transactions that do not know à priori the columns they will use (when using indexes, for example). Distributed per row locks - Key: CASSANDRA-2686 URL: https://issues.apache.org/jira/browse/CASSANDRA-2686 Project: Cassandra Issue Type: Wish Components: Core Environment: any Reporter: Luís Ferreira Labels: api-addition, features Instead of using a centralized locking strategy like cages with zookeeper, I would like to have it in a decentralized way. Even if it carries some limitations. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-2686) Distributed per row locks
[ https://issues.apache.org/jira/browse/CASSANDRA-2686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13037965#comment-13037965 ] Luís Ferreira commented on CASSANDRA-2686: -- Is there any way to do this? Can a consensus be achieved between the nodes responsible for a key? Maybe by creating a new kind of column that would have write once properties, and by using something like hinted handoff of locks if a node is down at the time of locking. I am going to try to implement this. But it would be really good to get some ideas on how to do it. Distributed per row locks - Key: CASSANDRA-2686 URL: https://issues.apache.org/jira/browse/CASSANDRA-2686 Project: Cassandra Issue Type: Wish Components: Core Affects Versions: 0.7.4 Environment: any Reporter: Luís Ferreira Labels: api-addition, features Instead of using a centralized locking strategy like cages with zookeeper, I would like to have it in a decentralized way. Even if it carries some limitations. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-2686) Distributed per row locks
[ https://issues.apache.org/jira/browse/CASSANDRA-2686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13038060#comment-13038060 ] Stu Hood commented on CASSANDRA-2686: - The Spinnaker paper described a complete solution to this problem that embedded Zookeeper: www.vldb.org/pvldb/vol4/p243-rao.pdf Distributed per row locks - Key: CASSANDRA-2686 URL: https://issues.apache.org/jira/browse/CASSANDRA-2686 Project: Cassandra Issue Type: Wish Components: Core Environment: any Reporter: Luís Ferreira Labels: api-addition, features Instead of using a centralized locking strategy like cages with zookeeper, I would like to have it in a decentralized way. Even if it carries some limitations. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-2686) Distributed per row locks
[ https://issues.apache.org/jira/browse/CASSANDRA-2686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13038071#comment-13038071 ] Luís Ferreira commented on CASSANDRA-2686: -- Yes, but this uses a centralized or semi centralized approach, in which zookeeper controls who has the lock. I would like for all (up) the nodes to reach an agreement on who has the lock. I don't know if this is possible, and if so, at what price... Distributed per row locks - Key: CASSANDRA-2686 URL: https://issues.apache.org/jira/browse/CASSANDRA-2686 Project: Cassandra Issue Type: Wish Components: Core Environment: any Reporter: Luís Ferreira Labels: api-addition, features Instead of using a centralized locking strategy like cages with zookeeper, I would like to have it in a decentralized way. Even if it carries some limitations. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-2686) Distributed per row locks
[ https://issues.apache.org/jira/browse/CASSANDRA-2686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13038070#comment-13038070 ] Ryan King commented on CASSANDRA-2686: -- You'll likely end up reimplementing something like Paxos (what google's chubby uses) or ZAB (what Zookeeper uses). Distributed per row locks - Key: CASSANDRA-2686 URL: https://issues.apache.org/jira/browse/CASSANDRA-2686 Project: Cassandra Issue Type: Wish Components: Core Environment: any Reporter: Luís Ferreira Labels: api-addition, features Instead of using a centralized locking strategy like cages with zookeeper, I would like to have it in a decentralized way. Even if it carries some limitations. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-2686) Distributed per row locks
[ https://issues.apache.org/jira/browse/CASSANDRA-2686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13038078#comment-13038078 ] Ryan King commented on CASSANDRA-2686: -- Those protocols are methods for reach[ing] agreement. You're basically describing how ZK works. Distributed per row locks - Key: CASSANDRA-2686 URL: https://issues.apache.org/jira/browse/CASSANDRA-2686 Project: Cassandra Issue Type: Wish Components: Core Environment: any Reporter: Luís Ferreira Labels: api-addition, features Instead of using a centralized locking strategy like cages with zookeeper, I would like to have it in a decentralized way. Even if it carries some limitations. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-2686) Distributed per row locks
[ https://issues.apache.org/jira/browse/CASSANDRA-2686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13038090#comment-13038090 ] Luís Ferreira commented on CASSANDRA-2686: -- Maybe I misunderstood, but to get locks with ZK, every node has to make it's presence known to ZK, and ask ZK to get a certain lock (which I think it's what cages does). Can't this be a bottleneck? I know consensus can't be achieved with one failing node, therefore the need for something like ZK. Still, isn't there a way to do this using something like hinted handoff? Maybe I haven't explained correctly my idea. I'd like to have locks, but maintain the general structure of a cassandra cluster, and change as little as possible the kind of messages the nodes send. Distributed per row locks - Key: CASSANDRA-2686 URL: https://issues.apache.org/jira/browse/CASSANDRA-2686 Project: Cassandra Issue Type: Wish Components: Core Environment: any Reporter: Luís Ferreira Labels: api-addition, features Instead of using a centralized locking strategy like cages with zookeeper, I would like to have it in a decentralized way. Even if it carries some limitations. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-2686) Distributed per row locks
[ https://issues.apache.org/jira/browse/CASSANDRA-2686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13038093#comment-13038093 ] Luís Ferreira commented on CASSANDRA-2686: -- Probably you're right, and the only way to do this is by using some kind of protocol as ZAB or Paxos. Still, if other ideas come up it would be great. Or even a explanation of why that's the only way... Distributed per row locks - Key: CASSANDRA-2686 URL: https://issues.apache.org/jira/browse/CASSANDRA-2686 Project: Cassandra Issue Type: Wish Components: Core Environment: any Reporter: Luís Ferreira Labels: api-addition, features Instead of using a centralized locking strategy like cages with zookeeper, I would like to have it in a decentralized way. Even if it carries some limitations. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-2686) Distributed per row locks
[ https://issues.apache.org/jira/browse/CASSANDRA-2686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13038107#comment-13038107 ] Jonathan Ellis commented on CASSANDRA-2686: --- Pretty sure Spinnaker only used ZK for leader election, not normal row updates. Distributed per row locks - Key: CASSANDRA-2686 URL: https://issues.apache.org/jira/browse/CASSANDRA-2686 Project: Cassandra Issue Type: Wish Components: Core Environment: any Reporter: Luís Ferreira Labels: api-addition, features Instead of using a centralized locking strategy like cages with zookeeper, I would like to have it in a decentralized way. Even if it carries some limitations. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (CASSANDRA-2686) Distributed per row locks
[ https://issues.apache.org/jira/browse/CASSANDRA-2686?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13038139#comment-13038139 ] Stu Hood commented on CASSANDRA-2686: - bq. Pretty sure Spinnaker only used ZK for leader election, not normal row updates. If you've done leader election such that one replica is the leader of the replica set, then there is no need for locking... although CAS operations must speak directly to the leader. Distributed per row locks - Key: CASSANDRA-2686 URL: https://issues.apache.org/jira/browse/CASSANDRA-2686 Project: Cassandra Issue Type: Wish Components: Core Environment: any Reporter: Luís Ferreira Labels: api-addition, features Instead of using a centralized locking strategy like cages with zookeeper, I would like to have it in a decentralized way. Even if it carries some limitations. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira
