Identical node name for all cluster nodes


(Vikram Adukia) #1

I just set up a 3 node elasticsearch cluster with version 0.18.7. For
some reason, all the nodes are coming up with the node name of
"Aardwolf".

All nodes are able to find each other. All nodes are using the same
cluster.name

Currently, there is nothing indexed in the cluster.

Currently, I don't have node.name set in the configuration file.
Though, if I did set it, it does work. However, without that setting,
no matter how many times I restart different nodes, the node name
always comes up as "Aardwolf".

Each node does have a unique id.

So, here are my questions:

  1. why is this happening? I don't see this behavior with another
    elasticsearch cluster that I have set up
  2. would this cause any problems?

thanks


(Igor Motov) #2

My guess would be that you are running Java 7u1 on this new cluster and
hitting known bug in ThreadLocalRandom class
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6955840. This bug is
fixed in 7u2, so you can ether upgrade java to 7u2 or 7u3 or downgrade it
to 1.6. Since ThreadLocalRandom is used in several other places in
elasticsearch, this bug might be causing other issues, but I don't know for
sure.

On Mon, Feb 27, 2012 at 7:47 PM, Vikram Adukia adukia@fireitup.net wrote:

I just set up a 3 node elasticsearch cluster with version 0.18.7. For
some reason, all the nodes are coming up with the node name of
"Aardwolf".

All nodes are able to find each other. All nodes are using the same
cluster.name

Currently, there is nothing indexed in the cluster.

Currently, I don't have node.name set in the configuration file.
Though, if I did set it, it does work. However, without that setting,
no matter how many times I restart different nodes, the node name
always comes up as "Aardwolf".

Each node does have a unique id.

So, here are my questions:

  1. why is this happening? I don't see this behavior with another
    elasticsearch cluster that I have set up
  2. would this cause any problems?

thanks


(Vikram Adukia-2) #3

Updating to 7u3 does appear to have fixed the issue.

thanks

On Mon, Feb 27, 2012 at 5:46 PM, Igor Motov imotov@gmail.com wrote:

My guess would be that you are running Java 7u1 on this new cluster and
hitting known bug in ThreadLocalRandom class
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6955840. This bug is
fixed in 7u2, so you can ether upgrade java to 7u2 or 7u3 or downgrade it
to 1.6. Since ThreadLocalRandom is used in several other places in
elasticsearch, this bug might be causing other issues, but I don't know for
sure.

On Mon, Feb 27, 2012 at 7:47 PM, Vikram Adukia adukia@fireitup.netwrote:

I just set up a 3 node elasticsearch cluster with version 0.18.7. For
some reason, all the nodes are coming up with the node name of
"Aardwolf".

All nodes are able to find each other. All nodes are using the same
cluster.name

Currently, there is nothing indexed in the cluster.

Currently, I don't have node.name set in the configuration file.
Though, if I did set it, it does work. However, without that setting,
no matter how many times I restart different nodes, the node name
always comes up as "Aardwolf".

Each node does have a unique id.

So, here are my questions:

  1. why is this happening? I don't see this behavior with another
    elasticsearch cluster that I have set up
  2. would this cause any problems?

thanks


(Shay Banon) #4

Note also that you should only use Java 7u2 and above, otherwise, you might get index corruptions due to a bug in Java 7 and 7u1 versions.

On Tuesday, February 28, 2012 at 6:22 PM, Vikram Adukia wrote:

Updating to 7u3 does appear to have fixed the issue.

thanks

On Mon, Feb 27, 2012 at 5:46 PM, Igor Motov <imotov@gmail.com (mailto:imotov@gmail.com)> wrote:

My guess would be that you are running Java 7u1 on this new cluster and hitting known bug in ThreadLocalRandom class http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6955840. This bug is fixed in 7u2, so you can ether upgrade java to 7u2 or 7u3 or downgrade it to 1.6. Since ThreadLocalRandom is used in several other places in elasticsearch, this bug might be causing other issues, but I don't know for sure.

On Mon, Feb 27, 2012 at 7:47 PM, Vikram Adukia <adukia@fireitup.net (mailto:adukia@fireitup.net)> wrote:

I just set up a 3 node elasticsearch cluster with version 0.18.7. For
some reason, all the nodes are coming up with the node name of
"Aardwolf".

All nodes are able to find each other. All nodes are using the same
cluster.name (http://cluster.name)

Currently, there is nothing indexed in the cluster.

Currently, I don't have node.name (http://node.name) set in the configuration file.
Though, if I did set it, it does work. However, without that setting,
no matter how many times I restart different nodes, the node name
always comes up as "Aardwolf".

Each node does have a unique id.

So, here are my questions:

  1. why is this happening? I don't see this behavior with another
    elasticsearch cluster that I have set up
  2. would this cause any problems?

thanks


(system) #5