cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sam Tunnicliffe (JIRA)" <>
Subject [jira] [Updated] (CASSANDRA-11038) Is node being restarted treated as node joining?
Date Fri, 20 May 2016 16:45:13 GMT


Sam Tunnicliffe updated CASSANDRA-11038:
    Fix Version/s: 3.x
           Status: Patch Available  (was: Open)

Pushed branches with fixes for 2.2/3.0/3.7/trunk - though the fix merges forward cleanly except
for conflicts where I've cleaned up imports. Basically, these preserve the existing behaviour
of delivering both {{NEW_NODE}} and {{UP}} events when a node first joins the cluster &
of delaying both until after the node becomes available for clients. The erroneous {{NEW_NODE}}
when a known node is restarted has been removed. The tracking of pushed notifications in {{EventNotifier}}
is still necessary at the moment (because [reasons|]),
but they will go away with CASSANDRA-9156. See CASSANDRA-11731 for some related discussion.

dtest branch [here|]


(so far I've only kicked off CI for the 2.2 branch, just in case there's some problem I didn't
run into locally, will kick off the other jobs when that finishes).

> Is node being restarted treated as node joining?
> ------------------------------------------------
>                 Key: CASSANDRA-11038
>                 URL:
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Distributed Metadata
>            Reporter: cheng ren
>            Assignee: Sam Tunnicliffe
>            Priority: Minor
>             Fix For: 2.2.x, 3.0.x, 3.x
> Hi, 
> What we found recently is that every time we restart a node, all other nodes in the cluster
treat the restarted node as a new node joining and issue node joining notification to clients.
We have traced the code path being hit when a peer node detected a restarted node:
> src/java/org/apache/cassandra/gms/
> {code}
>     private void handleMajorStateChange(InetAddress ep, EndpointState epState)
>     {
>         if (!isDeadState(epState))
>         {
>             if (endpointStateMap.get(ep) != null)
>       "Node {} has restarted, now UP", ep);
>             else
>       "Node {} is now part of the cluster", ep);
>         }
>         if (logger.isTraceEnabled())
>             logger.trace("Adding endpoint state for " + ep);
>         endpointStateMap.put(ep, epState);
>         // the node restarted: it is up to the subscriber to take whatever action is
>         for (IEndpointStateChangeSubscriber subscriber : subscribers)
>             subscriber.onRestart(ep, epState);
>         if (!isDeadState(epState))
>             markAlive(ep, epState);
>         else
>         {
>             logger.debug("Not marking " + ep + " alive due to dead state");
>             markDead(ep, epState);
>         }
>         for (IEndpointStateChangeSubscriber subscriber : subscribers)
>             subscriber.onJoin(ep, epState);
>     }
> {code}
> subscriber.onJoin(ep, epState) ends up with calling onJoinCluster in
> {code}
> src/java/org/apache/cassandra/transport/
>         public void onJoinCluster(InetAddress endpoint)
>         {
> server.connectionTracker.send(Event.TopologyChange.newNode(getRpcAddress(endpoint), server.socket.getPort()));
>         }
> {code}
> We have a full trace of code path and skip some intermedia function calls here for being
> Upon receiving the node joining notification, clients would go and scan system peer table
to fetch the latest topology information. Since we have tens of thousands of client connections,
scans from all of them put an enormous load to our cluster. 
> Although in the newer version of driver, client skips fetching peer table if the new
node has already existed in local metadata, we are still curious why node being restarted
is handled as node joining on server side? Did we hit a bug or this is the way supposed to
be? Our old java driver version is 1.0.4 and cassandra version is 2.0.12.
> Thanks!

This message was sent by Atlassian JIRA

View raw message