Hey,
First, it seems like all the logging is DEBUG mode, thats noisy, all you need is discovery set to TRACE.
If you see, on both nodes, the output: "using dynamic discovery nodes" is empty, meaning that it went ahead and queries aws for the list of nodes, and got none... . Maybe the groups / tags are not correct?
-shay.banon
On Thursday, July 7, 2011 at 3:10 PM, Ridvan Gyundogan wrote:
Shay,
here is the gist:
master · GitHubOn Jul 6, 10:35 pm, Shay Banon <shay.ba...@elasticsearch.com (http://elasticsearch.com)> wrote:
Can you set logging for discovery to TRACE in the logging file and gist it?
On Wednesday, July 6, 2011 at 9:35 PM, Ridvan Gyundogan wrote:
Thanks Shay.
Now it is bit better, at least I do not have an exception but still
the ec2 instances do not run as a cluster.
What I additionally did is the following. Enabled ping for the
security group with the ec2-authorise. Tried that the 2 machines can
ping each other.
Checked with telnet that the 2 machines can open each other's 9200 and
9300 ports. The 9301 - 9303 ports are open too but the other ports are
closed by default.
These are the 2 log files I have. First the master and next the one
which had to become a slave.At the bottom is the elasticsearch.yml
config file I use.
Any ideas what I need to change?master:
[2011-07-06 18:14:42,458][INFO ][node ] [Jimmy
Woo] {elasticsearch/0.16.2}[2142]: initializing ...
[2011-07-06 18:14:42,468][INFO ][plugins ] [Jimmy
Woo] loaded [cloud-aws]
[2011-07-06 18:14:48,272][INFO ][node ] [Jimmy
Woo] {elasticsearch/0.16.2}[2142]: initialized
[2011-07-06 18:14:48,272][INFO ][node ] [Jimmy
Woo] {elasticsearch/0.16.2}[2142]: starting ...
[2011-07-06 18:14:48,414][INFO ][transport ] [Jimmy
Woo] bound_address {inet[/10.224.50.50:9300]}, publish_address {inet[/
10.224.50.50:9300]}
[2011-07-06 18:14:52,797][INFO ][cluster.service ] [Jimmy
Woo] new_master [Jimmy Woo][T_0Y3f_UQJmfuiqYYW7Ivg][inet[/
10.224.50.50:9300]], reason: zen-disco-join (elected_as_master)
[2011-07-06 18:14:52,845][INFO ][discovery ] [Jimmy
Woo] perftests/T_0Y3f_UQJmfuiqYYW7Ivg
[2011-07-06 18:14:52,893][INFO ][gateway ] [Jimmy
Woo] recovered [0] indices into cluster_state
[2011-07-06 18:14:52,894][INFO ][http ] [Jimmy
Woo] bound_address {inet[/10.224.50.50:9200]}, publish_address {inet[/
10.224.50.50:9200]}
[2011-07-06 18:14:52,895][INFO ][node ] [Jimmy
Woo] {elasticsearch/0.16.2}[2142]: startedslave:
[2011-07-06 18:18:07,733][INFO ][node ] [Machine
Teen] {elasticsearch/0.16.2}[1251]: initializing ...
[2011-07-06 18:18:07,741][INFO ][plugins ] [Machine
Teen] loaded [cloud-aws]
[2011-07-06 18:18:11,994][INFO ][node ] [Machine
Teen] {elasticsearch/0.16.2}[1251]: initialized
[2011-07-06 18:18:11,995][INFO ][node ] [Machine
Teen] {elasticsearch/0.16.2}[1251]: starting ...
[2011-07-06 18:18:12,105][INFO ][transport ] [Machine
Teen] bound_address {inet[/10.234.227.72:9300]}, publish_address
{inet[/10.234.227.72:9300]}
[2011-07-06 18:18:16,242][INFO ][cluster.service ] [Machine
Teen] new_master [Machine Teen][4vSbOGJqRx6q3Vi5HrJ_IA][inet[/
10.234.227.72:9300]], reason: zen-disco-join (elected_as_master)
[2011-07-06 18:18:16,278][INFO ][discovery ] [Machine
Teen] perftests/4vSbOGJqRx6q3Vi5HrJ_IA
[2011-07-06 18:18:16,305][INFO ][gateway ] [Machine
Teen] recovered [0] indices into cluster_state
[2011-07-06 18:18:16,312][INFO ][http ] [Machine
Teen] bound_address {inet[/10.234.227.72:9200]}, publish_address
{inet[/10.234.227.72:9200]}
[2011-07-06 18:18:16,313][INFO ][node ] [Machine
Teen] {elasticsearch/0.16.2}[1251]: startedelasticsearch.yml:
cluster:
name: perftestsdiscovery:
type: ec2
zen:
ping_timeout: 30scloud:
account: ABCDEFGIJKLAAAAAAAAAA #here I use the "Access Key ID",
not the "Account Number"
key: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
compute:
type: amazonnetwork:
bind_host: eth0:ipv4On Jul 6, 7:16 pm, Shay Banon <shay.ba...@elasticsearch.com (http://elasticsearch.com)> wrote:
You need to set the discovery type to ec2:Elasticsearch Platform — Find real-time answers at scale | Elastic. The blog post is outdated... .
And you don't have to use master to have discovery working on ec2. You can set network.host to whatever value you want, for example, the internal IP address. If you want teh same config across nodes, you can bind to the network interface name (Elasticsearch Platform — Find real-time answers at scale | Elastic), for example: en0:ipv4.
On Wednesday, July 6, 2011 at 4:34 PM, Ridvan Gyundogan wrote:
Unfortunately using the trunk is not an option at the moment because
there are some maven dependencies to the 16.2 jar in our project.
We will upgrade to 0.17 once it is officially released.Now I tried the "Zen" as described here:
Elasticsearch Platform — Find real-time answers at scale | Elasticmy elasticsearch.yml on the slave looks like this:
cluster:
name: perftestsdiscovery:
type: cloud
zen:
ping_timeout: 30s
cloud:
account: 551234567890
key: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
compute:
type: amazon
network:
bind_host: non_loopback:ipv4On this I get the exception:
{elasticsearch/0.16.2}: Initialization Failed ...
- NoClassSettingsException[Failed to load class setting
[discovery.type] with value [cloud]]
ClassNotFoundException[cloud]for the network.bind_host I tried eth0:ipv4 but this gives me host
notfound exception.Any ideas? Can someone upload an elasticsearch.yml file able to do ec2
discovery for him with the 16.2 version?On Jul 6, 2:40 pm, Ridvan Gyundogan <ridva...@gmail.com (http://gmail.com)> wrote:
Ah ok thanks I will try.
On Wed, Jul 6, 2011 at 2:37 PM, Clinton Gormley <clin...@iannounce.co.uk (http://iannounce.co.uk)>wrote:
I start only 1 node at the moment so sure it should be the master.
By master, Shay means the version of Elasticsearch that is currently
master on github, not a released version.You need to download and compile it yourself:
Elasticsearch Platform — Find real-time answers at scale | Elastic
clint