[jboss-jira] [JBoss JIRA] (JGRP-1906) jdbc_ping doesn't delete crashed node
Bela Ban (JIRA)
issues at jboss.org
Mon Jan 19 09:45:49 EST 2015
[ https://issues.jboss.org/browse/JGRP-1906?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13033563#comment-13033563 ]
Bela Ban commented on JGRP-1906:
--------------------------------
{{JDBC_PING}} registers a shutdown hook when starting. When the process is killed or stopped, the process removes its information from the DB table. However, this is not done when you kill a process with kill -9. In this case, you have to manually remove the information.
The reason failure detection doesn't do this is because of split brains, e.g. if we have A,B,C,D split into AB and CD. In this case, A cannot remove C and D, and C cannot remove A and B, because the members are still alive.
> jdbc_ping doesn't delete crashed node
> -------------------------------------
>
> Key: JGRP-1906
> URL: https://issues.jboss.org/browse/JGRP-1906
> Project: JGroups
> Issue Type: Bug
> Affects Versions: 3.6.1
> Environment: n/a
> Reporter: rama rama
> Assignee: Bela Ban
> Priority: Critical
>
> Hi,
> i am having trouble with jgroups. If a node crash, it doesn't get removed from jdbc table.
> After 10 hours of work, the table contains 100 rows (that is not a problem) but, when the server start_up, GMS take _AGES_ to try to connect to all this dead node.
> Since i am only a 'user' speaking about jgroup and i have no idea on how internally does it work, it this normal?
> I don't think so, master of the cluster, or a governor or something like that, should be able to detect if a node is present or not (via FD,VERIFY_SUSPECT) and delete it from JDBC table to avoid issue on startup for the next time, correct me if i am wrong.
> Here a copy/paste of my current config, configured in applicative way.
> ------------------------------------------------------------------------------------
> int min_cores = 1;
> int max_cores = 50;
> InetAddress bind_addr = org.jgroups.util.Util.getAddressByPatternMatch("match-address:" + Config.get("Cluster.bind_addr"));
> stack
> .addProtocol(new UDP()
> .setValue("bind_addr", bind_addr)
> .setValue("loopback", true)
> .setValue("thread_naming_pattern", "cl")
> .setValue("timer_type", "new3")
> .setValue("timer_min_threads", min_cores)
> .setValue("timer_max_threads", max_cores)
> .setValue("timer_keep_alive_time", Util.MIN * 10)
> .setValue("timer_queue_max_size", 500)
> .setValue("mcast_port", Config.get("Cluster.mcast_port", Constant.INT_NULL))
> .setValue("thread_pool_keep_alive_time", Util.MIN * 10)
> .setValue("thread_pool_min_threads", min_cores)
> .setValue("thread_pool_max_threads", max_cores)
> .setValue("thread_pool_queue_enabled", false)
> .setValue("thread_pool_queue_max_size", 500)
> .setValue("oob_thread_pool_keep_alive_time", Util.MIN * 10)
> .setValue("oob_thread_pool_min_threads", min_cores)
> .setValue("oob_thread_pool_max_threads", max_cores)
> .setValue("oob_thread_pool_queue_enabled", false)
> .setValue("oob_thread_pool_queue_max_size", 500)
> .setValue("internal_thread_pool_keep_alive_time", Util.MIN * 10)
> .setValue("internal_thread_pool_min_threads", min_cores)
> .setValue("internal_thread_pool_max_threads", max_cores)
> .setValue("internal_thread_pool_queue_enabled", false)
> .setValue("internal_thread_pool_queue_max_size", 600000)
> .setValue("ucast_recv_buf_size", org.jgroups.util.Util.readBytesInteger("5M"))
> .setValue("ucast_send_buf_size", org.jgroups.util.Util.readBytesInteger("640K"))
> .setValue("mcast_recv_buf_size", org.jgroups.util.Util.readBytesInteger("5M"))
> .setValue("mcast_send_buf_size", org.jgroups.util.Util.readBytesInteger("640K"))
> )
> .addProtocol(new JDBC_PING()
> .setValue("connection_url", Config.get("DB.dbdriver") + ':' + Config.get("DB.dburl") + '/' + Config.get("DB.dbname"))
> .setValue("connection_username", Config.get("DB.dbuser"))
> .setValue("connection_password", Config.get("DB.dbpwd"))
> .setValue("connection_driver", "org.postgresql.Driver")
> .setValue("initialize_sql", "CREATE TABLE IF NOT EXISTS JGROUPSPING ( own_addr varchar(200) NOT NULL, cluster_name varchar(200) NOT NULL, ping_data bytea DEFAULT NULL, PRIMARY KEY (own_addr, cluster_name) )")
> )
> .addProtocol(new MERGE3()
> .setValue("max_interval", 10000)
> .setValue("min_interval", 1000)
> )
> .addProtocol(new FD_SOCK()
> .setValue("bind_addr", bind_addr)
> )
> .addProtocol(new FD_ALL()
> )
> .addProtocol(new VERIFY_SUSPECT()
> .setValue("timeout", 1000)
> .setValue("bind_addr", bind_addr)
> )
> .addProtocol(new NAKACK2()
> .setValue("xmit_interval", 500)
> .setValue("xmit_table_num_rows", 100)
> .setValue("xmit_table_msgs_per_row", 2000)
> .setValue("xmit_table_max_compaction_time", 30000)
> .setValue("max_msg_batch_size", 500)
> .setValue("use_mcast_xmit", false)
> .setValue("discard_delivered_msgs", true)
> )
> .addProtocol(new UNICAST3()
> .setValue("xmit_table_num_rows", 100)
> .setValue("xmit_table_msgs_per_row", 1000)
> .setValue("xmit_table_max_compaction_time", 30000)
> .setValue("max_msg_batch_size", 500)
> )
> .addProtocol(new STABLE()
> .setValue("stability_delay", 2000)
> .setValue("desired_avg_gossip", 60000)
> .setValue("max_bytes", org.jgroups.util.Util.readBytesInteger("4M"))
> .setValue("cap", 0.1)
> )
> .addProtocol(new GMS()
> .setValue("join_timeout", 3000)
> .setValue("view_bundling", false)
> .setValue("print_local_addr", false)
> .setValue("print_physical_addrs", false)
> )
> .addProtocol(new UFC()
> .setValue("max_credits", org.jgroups.util.Util.readBytesInteger("4M"))
> .setValue("min_threshold", 0.4)
> )
> .addProtocol(new MFC()
> .setValue("max_credits", org.jgroups.util.Util.readBytesInteger("4M"))
> .setValue("min_threshold", 0.4)
> )
> .addProtocol(new FRAG2()
> .setValue("frag_size", org.jgroups.util.Util.readBytesInteger("60K"))
> )
> .addProtocol(new RSVP()
> .setValue("resend_interval", 2000)
> .setValue("timeout", 10000)
> )
> ;
> stack.init();
> ----------------------------
--
This message was sent by Atlassian JIRA
(v6.3.11#6341)
More information about the jboss-jira
mailing list