Clebert-
Can you clarify what you mean by:
"..the server will shutdown.."
Are you suggesting a warm mode like "slave" where acceptors and cluster
connections are turned down, or a full shutdown of the process?
On 11/22/16 8:29 AM, Clebert Suconic wrote:
I could add a configuration for the ping command... The issue is with
parameters that may vary on different implementations.
I was providing the ping command already as a backup in case isReachable failed.
I will think about the IPV6. How I would write a test case for the IPV6? :/
Also, I have been thinking of better names for these configuration
options. if anyone has a more obvious name I would appreciate the
help.
Currently I have:
<network-check-NIC> - The network Interface Card name
<network-check-period> - The interval in MS
<network-check-timeout> - The timeout to fail the ping in Milliseconds
<network-check-URL-list> - Similar to checkList, but this will use
I may add a <network-check-ping-command>... and maybe use an
expression for the parameter.
On Tue, Nov 22, 2016 at 8:41 AM, Miroslav Novak <[email protected]> wrote:
Hi Clebert,
the idea looks ok.
Maybe it could be more system agnostic. I think that ping command might not be
present on some OS installations. InetAddress#isReachable(int) works in best
effort mode based on [1] and need some privileges. Do you think it could be
possible to provide some default implementation and allow user to add his own
kind of check of IP addresses.
Also this should support IPv6 addresses. wdyt?
Thanks,
Mirek
[1]
https://docs.oracle.com/javase/8/docs/api/java/net/InetAddress.html#isReachable-int-
----- Original Message -----
From: "Clebert Suconic" <[email protected]>
To: [email protected]
Sent: Tuesday, November 22, 2016 2:17:19 AM
Subject: [DISCUSS] ARTEMIS-863 - Network Health Check
I am implementing a feature that will allow an user to specify a ping
address list.
the broker will ping such list on a given interval, and it will
shutdown once the list cannot be achieved.
For instance, you specify 10.0.0.1, the broker will ping 10.0.0.1
every 10 seconds, and if that address is not reachable, it would go
down. (shutdown). this is to prevent a network isolation, so in case
you pull out the cable of the server the server will shutdown until
the network is back.
There are 4 properties currently implemented:
<network-check-list> - IPs to be pinged (or DNS) .. this is simple
ICMP through Network.isReachable
<network-check-NIC> - The network Interface Card name
<network-check-period> - The interval in MS
<network-check-timeout> - The timeout to fail the ping in Milliseconds
<network-check-URL-list> - Similar to checkList, but this will use
HTTP open in a similar fashion.
These will work nicely by setting System properties. (which I will
further improve by allowing something like ./artemis run
-Cnetwork-check-list=10.0.0.2)
I have already implemented this into a branch which I will send tomorrow as
PR.
https://github.com/clebertsuconic/activemq-artemis/commits/netcheck
(BTW this link will not be valid for future readers of the list.. on
that case... just look at the commits on master and look for
ARTEMIS-863 on the git logs).
--
Clebert Suconic
