[jboss-jira] [JBoss JIRA] (JGRP-2253) FD_SOCK is not working in AWS environment

[Sibin Karnavar (JIRA)]

    [ https://issues.jboss.org/browse/JGRP-2253?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13596483#comment-13596483 ] 

Sibin Karnavar commented on JGRP-2253:
--------------------------------------

In FD_SOCK,

startServerSocket() - Here we have an option to specify a start_port and and a port_range. If the specified start_port is already in use, we are trying up to the specified port_range to create the socket address.

Why we are not doing the same during setupPingSocket() ?

When an application specifies client_bind_port, Why don't FD_SOCK consider the port_range if the specified port is unavailable similar to startServerSocket()?

I have turned on debug logs of FD_SOCK and i was able to see that,

2018-06-25 15:01:15.067 DEBUG 6170 --- [FD_SOCK pinger-14,xx-SKS-test_SJX_210620181404_XSJ,ip-xx-xx-xx-xx-60384] org.jgroups.protocols.FD_SOCK            :  -  - ip-xx-xx-xx-xx-60384: failed connecting to ip-XX-xxx-xxx-47443: Address already in use (Bind failed)

In AWS environment's , it is not a good practice to define a wide range of inbound open ports. We are more comfortable to define a smaller set of port range as allowed inbound ports for an EC2 instance.

The above may not be a solution for the AWS EC2 termination but it is really good to have. I think,  as you have mentioned, termination on EC2/AWS may not be closing the socket. Because I have removed my client_bind_port and opened EC2 security group rules to allow any inbound ports for testing purpose but FD_SOCK was still not working as expected. and the DEBUG/TRACE logs were clean with no issues


Thanks,
Sibin 












> FD_SOCK is not working in AWS environment
> -----------------------------------------
>
>                 Key: JGRP-2253
>                 URL: https://issues.jboss.org/browse/JGRP-2253
>             Project: JGroups
>          Issue Type: Bug
>    Affects Versions: 4.0.10
>         Environment: AWS - EC2
>            Reporter: Sibin Karnavar
>            Assignee: Bela Ban
>             Fix For: 4.0.13
>
>
> We have our failure detection defined like below. 
>  <FD_SOCK  external_port="7804" />
>  <FD timeout="3000" max_tries="3" />
> <VERIFY_SUSPECT timeout="3000" />
> Please note that we have used FD instead of FD_ALL in AWS. We will be changing it to FD_ALL later after detailed testing.
> In my local, this is working perfect. As soon as I kill my node, I was able to see that view change was happening immediately with FD_SOCK.
> We were not mentioning the external_port in the FD_SOCK but later I thought it may be an issue with the port and defined it as 7804 and added the same port to the security group that allows to access this port among all the nodes.  So no issue with the port.
> Can you please let us know if we need any additional configurations to make FD_SOCK works well in AWS.
> Thanks,
> Sibin



--
This message was sent by Atlassian JIRA
(v7.5.0#75005)