Can't hit API on non-loopback IP

Elastic Stack Elasticsearch
1

I have just setup a new dev 7.6.2 single-node cluster via rpm on centos 7 and am unable to curl the API on the non-loopback address. Curl to localhost works fine and I can also telnet to the default port of 9200 of the non-loopback address from another host. I have tried many different variations for network host including special site, ip addr, 0.0.0.0, and 0. This is a POC so I am not trying to configure very much beyond defaults. I have no issues with the service starting in regards to the bootstrap checks for single node cluster. Any help would be much appreciated.

Relevant elasticsearch.yml network/http settings I have explicitly set in file:
network.host: 0
http.port: 9200

2

Try this in elasticsearch.yml:
network.host: ["ipAddress"]
discovery.type: single-node

And restart the elastic service

3

ipAddress is the public IP address of the device which you have set up ES.

4

Thank you for your reply, but unfortunately this did not fix the communication issue. Additionally I have no issue with the service starting as a single node cluster and electing itself as master.

5

Can you share what is the error are you getting?

6

I receive no error as far as the service goes. The issue is I am unable to communicate/receive a response from the api on the non-loopback address via curl on port 9200, despite the server being bound to said port and also being able to successfully telnet to 9200 from a different host as well.

7

When the service starts can you see what is the publish address in the logs?

8

yes, and it looks correct. From log file:

publish_address {10.240.10.221:9200}, bound_addresses {0.0.0.0:9200}

9

if you put network.host:["10.240.10.221"] in yml file , I think bound address should be 10.240.10.221:9200.
Have you assigned 0 to network.host?

10

I just tried your suggestion, and still am unable to get a response via curl.
From log file:
publish_address {10.240.10.221:9200}, bound_addresses {10.240.10.221:9200}

I also tried not using 0, and using [local, "10.240.10.221"]
From log file:
publish_address {10.240.10.221:9200}, bound_addresses {127.0.0.1:9200}, {10.240.10.221:9200}

Same result, works with loopback but not with ip addr and I can successfully telnet from another host on 10.240.10.221:9200. so it doesn't seem to be comm/firewall issue. Very confused!

11

use only ["10.240.10.221"]. It should work according to me.
Then you can only hit to "10.240.10.221" but not localhost, but it should be fine right?

12

yes, that is what I tried first. From the above, with no luck.

13

Ok. Can you share the curl call?

14

hitting localhost:
curl localhost:9200 - This works
trying from ES host itself and another host(can telnet over 9200 from this host):
curl 10.240.10.221:9200 - doesn't work

15

Looks fine to me. May be @dadoonet can help

16

Did you check if there is a firewall or something?

17

Both of these hosts have firewalld disabled for testing. They are two EC2 instances in the same security group with these ports open and I have successfully been able to telnet over port 9200 to the elasticsearch host from the other host. I have not configured the EC2 discovery plugin, although I may be incorrect seeing it as optional. Am I missing something that would be specific to it being cloud servers?

18

Are you saying that you can telnet IP:9200 but not curl?

19

@dadoonet Correct

20

Once you are connected with telnet can you run a GET / ?