EOFException during discovery


(Wojciech Durczyński) #1

Hello

I have two ES clusters (different cluster names) in my network. Nodes
discovery is via multicast,
Both clusters was version 0.17.8 but I upgraded one of them to use ES
0.18.6.
And now upgraded cluster nodes log this exception very often (log files
grow very fast):

16:28:57,744 WARN [cast#receiver-pool-16-thread-1]
asticsearch.discovery.zen.ping.multicast [,,,] - [Madame Menace] failed to
read requesting node from /10.133.105.131:54328
java.io.EOFException
at
org.elasticsearch.common.io.stream.BytesStreamInput.readByte(BytesStreamInput.java:92)
at
org.elasticsearch.common.io.stream.HandlesStreamInput.readByte(HandlesStreamInput.java:80)
at
org.elasticsearch.common.io.stream.StreamInput.readVInt(StreamInput.java:74)
at org.elasticsearch.Version.readVersion(Version.java:55)
at
org.elasticsearch.cluster.node.DiscoveryNode.readFrom(DiscoveryNode.java:221)
at
org.elasticsearch.cluster.node.DiscoveryNode.readNode(DiscoveryNode.java:207)
at
org.elasticsearch.discovery.zen.ping.multicast.MulticastZenPing$Receiver.run(MulticastZenPing.java:364)
at java.lang.Thread.run(Thread.java:619)

Of course disabling or upgrading second cluster solves problem but why is
it impossible to maintain two clusters with different versions and cluster
names?
How to disable this warning in logging configuration?

Regards
Wojciech Durczyński


(Shay Banon) #2

It is possible to maintain two different versions, its just the exception
that is annoying, it might make sense to change the log level on this
one... .

2012/1/9 Wojciech Durczyński wojciech.durczynski@comarch.com

Hello

I have two ES clusters (different cluster names) in my network. Nodes
discovery is via multicast,
Both clusters was version 0.17.8 but I upgraded one of them to use ES
0.18.6.
And now upgraded cluster nodes log this exception very often (log files
grow very fast):

16:28:57,744 WARN [cast#receiver-pool-16-thread-1]
asticsearch.discovery.zen.ping.multicast [,,,] - [Madame Menace] failed to
read requesting node from /10.133.105.131:54328
java.io.EOFException
at
org.elasticsearch.common.io.stream.BytesStreamInput.readByte(BytesStreamInput.java:92)
at
org.elasticsearch.common.io.stream.HandlesStreamInput.readByte(HandlesStreamInput.java:80)
at
org.elasticsearch.common.io.stream.StreamInput.readVInt(StreamInput.java:74)
at org.elasticsearch.Version.readVersion(Version.java:55)
at
org.elasticsearch.cluster.node.DiscoveryNode.readFrom(DiscoveryNode.java:221)
at
org.elasticsearch.cluster.node.DiscoveryNode.readNode(DiscoveryNode.java:207)
at
org.elasticsearch.discovery.zen.ping.multicast.MulticastZenPing$Receiver.run(MulticastZenPing.java:364)
at java.lang.Thread.run(Thread.java:619)

Of course disabling or upgrading second cluster solves problem but why is
it impossible to maintain two clusters with different versions and cluster
names?
How to disable this warning in logging configuration?

Regards
Wojciech Durczyński


(Wojciech Durczyński) #3

How to do this? Something like:
org.elasticsearch.discovery.zen.ping.multicast: ERROR
in logging.yml? Won't it mute many other interesting logs?


(Shay Banon) #4

Yea, I meant it might make sense to change the level that this failure is
logged under..., though, it does help when trying to nail down discovery
problems with mixed versions (which is why I put it under WARN to begin
with)..

2012/1/9 Wojciech Durczyński wojciech.durczynski@comarch.com

How to do this? Something like:
org.elasticsearch.discovery.**zen.ping.multicast: ERROR
in logging.yml? Won't it mute many other interesting logs?


(system) #5