ArithmeticException: / by zero in ClusterConnectionSelector (#1266)

Activities

wojciech.kapcia@tigase.net opened 3 years ago

[2021-07-26 06:25:31:329] [SEVERE  ] [         in_14-cl-comp ] AbstractMessageReceiver$QueueListener.run(): [in_14-cl-comp] Exception during packet processing: from=null, to=null, DATA=<cluster pr="HIGH" type="set" xmlns="tigase:cluster" id="cl-1680672" to="sess-man@ip-172-31-20-109.us-west-2.compute.internal" from="sess-man@ip-172-31-0-99.us-west-2.compute.internal"><control><visited-nodes><node-id>sess-man@ip-172-31-0-99.us-west-2.compute.internal</node-id></visited-nodes><method-call name="sess-man-packet-forward-sm-cmd"><par name="stable-id">985d9f45-a0ba-42c4-932c-7d60b8ecbe5c</par></method-call><first-node>sess-man@ip-172-31-0-99.us-west-2.compute.internal</first-node></control>…, SIZE=1238, XMLNS=tigase:cluster, PRIORITY=HIGH, PERMISSION=NONE, TYPE=set, STABLE_ID=null
java.lang.ArithmeticException: / by zero
	at tigase.cluster.ClusterConnectionSelector.selectConnection(ClusterConnectionSelector.java:67)
	at tigase.cluster.ClusterConnectionManager.writePacketToSocket(ClusterConnectionManager.java:813)
	at tigase.cluster.ClusterConnectionManager.processPacket(ClusterConnectionManager.java:401)
	at tigase.server.AbstractMessageReceiver$QueueListener.run(AbstractMessageReceiver.java:1404)

wojciech.kapcia@tigase.net commented 3 years ago

Most likely caused by volatile state of the cluster (one of the nodes was completely non-operational)

wojciech.kapcia@tigase.net commented 3 years ago

Concurrency issue: trying to get the connection to send the packet while it was removed. We could try catching this exception and returning null but that would would only change where the Exception is handled (ClusterConnectionSelector or AMR).

Type	Bug
Priority	Normal
Assignee	wojciech.kapcia@tigase.net
Version	tigase-server-8.2.0
Spent time	1h

Issue Votes (0)

Watchers (0)

Reference

tigase/_server/server-core#1266