]
Bela Ban edited comment on JGRP-1855 at 7/31/14 2:17 AM:
---------------------------------------------------------
OK, first impl is done and has been tested with VirtualBox/Fedora 14 instances. The config
would look like this:
{code:xml}
<FD_HOST interval="10000" timeout="35000" />
{code}
The current impl uses {{InetAddress.isReachable()}}. Next: include ability to use external
commands such as {{/sbin/ping}}.
was (Author: belaban):
OK, first impl is done and has been tested with VirtualBox/Fedora 14 instances. The config
would look like this:
{code:xml}
<FD_PING2 interval="10000" timeout="35000" />
{code}
The current impl uses {{InetAddress.isReachable()}}. Next: include ability to use external
commands such as {{/sbin/ping}}.
FD_HOST: host failure detection protocol
----------------------------------------
Key: JGRP-1855
URL:
https://issues.jboss.org/browse/JGRP-1855
Project: JGroups
Issue Type: Feature Request
Security Level: Public(Everyone can see)
Reporter: Bela Ban
Assignee: Bela Ban
Fix For: 3.4.5, 3.5
A new protocol similar to FD_PING which detects entire host failures and suspects all
members on the failed host.
Features are:
* Contrary to FD_PING which uses a ring structure, FD_HOST will have everyone ping
everybody else (similar to FD_ALL)
** A structure keeps track of hosts (IP addresses) and members on those hosts
*** Example
||192.168.1.2||192.168.1.3||192.168.1.5||
|A,B,C|D,E,F|X,Y,Z|
* We sort the members lexically and the *first* member runs a ping against each other IP
address, e.g. A pings 192.168.1.3 and 192.168.1.5, D pings 192.168.1.2 and 192.168.1.5
etc
* The ping command itself is pluggable and can be a Java class (e.g. using
{{InetAddress.isReachable()}}, a script or a command (e.g. {{/sbin/ping}}).
* When an entire host is suspected, we suspect *all* cluster members on it
** Example: if B suspects 192.168.1.5, members X, Y and Z are suspected and removed from
the view