Connection Refused


(newsgroup@syslog.com) #1

I'm getting the following errors:

[2012-04-16 13:17:31,897][INFO ][discovery.zen ] [app2-uk-
example-com] failed to send join request to master [[app2-prod-example-
com][T8eeWM5xTJ-cJw6FhF5Qxg][inet[/172.16.10.2:9300]]{zone=to}],
reason [org.elasticsearch.transport.RemoteTransportException: [app2-
prod-example-com][inet[/172.16.10.2:9300]][discovery/zen/join];
org.elasticsearch.transport.NotSerializableTransportException:
[org.elasticsearch.transport.ConnectTransportException] [app2-uk-
example-com][inet[/172.28.5.3:9300]] connect_timeout[30s]; Connection
refused; ]

I have two zones, to and uk, with the following config:

cluster.name: prod
node.name: ${NODENAME}
path.logs: /var/log/elasticsearch
network.host: 0.0.0.0
http.port: 9200
discovery.zen.minimum_master_nodes: 2
discovery.zen.ping.unicast.hosts: ["app1.prod", "app2.prod",
"app3.prod", "app4.prod", "app5.prod"]
node.zone: to
cluster.routing.allocation.awareness.attributes: zone

But the UK can't join anymore and I don't know why.

What I want is to have each zone have 2 copies of the index, and I
have four zones total. With the zone restriction, I was getting one
copy and I'd be happy with that again if I can get the UK server to
connect.


(newsgroup@syslog.com) #2

Further into this, I've removed all references to the cluster routing
and gone back to my original config, and the cluster master is still
refusing connection to the remote server.

I can telnet to port 9300 on the master but I don't get any response.
This is version 0.18.6.


(newsgroup@syslog.com) #3

We're on to day two of this being broken in our production
environment. I upgraded the entire cluster to 0.19.2 and it started
working, but then last night the UK server disconnected and it hasn't
been able to re-connect since.

Is there no one who can help me with this problem?


(Shairon Toledo) #4

We have a similar issue running ES over EC2 + Ubuntu 10.04.1 LTS, the fix
was migrating ubuntu to version 11.10.

On Tue, Apr 17, 2012 at 8:42 AM, newsgroup@syslog.com
newsgroup@syslog.comwrote:

We're on to day two of this being broken in our production
environment. I upgraded the entire cluster to 0.19.2 and it started
working, but then last night the UK server disconnected and it hasn't
been able to re-connect since.

Is there no one who can help me with this problem?

--

Shairon Toledo
http://hashcode.me


(newsgroup@syslog.com) #5

On Apr 17, 2:27 pm, Shairon Toledo shairon.tol...@gmail.com wrote:

We have a similar issue running ES over EC2 + Ubuntu 10.04.1 LTS, the fix
was migrating ubuntu to version 11.10.

I'm running CentOS 5.8.


(newsgroup@syslog.com) #6

Here's an update on this issue. It turns out that elasticsearch, when
configured to listen on any address, would always select the aliased
IP of the box in question. Manually setting the IP to the primary
interface has fixed this issue.

So, for future Googlers, if you have multiple IPs on the box where
you're running ES, manually configure the interface.


(newsgroup@syslog.com) #7

Here's an update on this issue. It turns out that elasticsearch, when
configured to listen on any address, would always select the aliased
IP of the box in question. Manually setting the IP to the primary
interface has fixed this issue.

So, for future Googlers, if you have multiple IPs on the box where
you're running ES, manually configure the interface.


(system) #8