Hi JB,
I quit using cellar-cloud because using a blob store (S3 in my case) left a lot
of clutter to clean up and didn't look very efficient. I would regard that as a
solution of last resort.
With AWS, tagging machines and scanning them with the describe-instances API is
more direct (no need to signIn/signOut or clean up the whiteboard when a
machine is terminated).
As to what I did, I'm not sure it's generic enough a solution at this stage
because I'm depending on : mvn: org.apache.servicemix.bundles /
org.apache.servicemix.bundles.aws-java-sdk / 1.8.9.1_1 because I'm using
ServiceMix, and I don't know what to depend on if I was to write a patch for
Karaf: jclouds perhpaps? Using plain AWS REST API would be an option (Hazelcast
folks do that) but involve some more work I can't put in just now.
Also should be noted that, from my experience, you need to disable the
automatic port-increment setting in hazelcast.xml otherwise it won't work! I
suspect AWS watch out for port scanning, in any case the cluster fails to form.
I used the Kubernetes as a template, which works exactly the same except for
the API used. So:
* implement the ManagedServiceFactory (so you can scan for multiple tags/labels)
* create your API request beforehand
* do the following
public void init() {
try {
if (amazonEC2 == null) {
amazonEC2 = new AmazonEC2Client();
}
amazonEC2.setRegion(region);
describeInstancesRequest = createRequest();
LOGGER.debug("CELLAR EC2: discovery service initialized");
} catch (Exception e) {
LOGGER.error("CELLAR EC2: can't init discovery service for
region " + region, e);
}
}
@Override
public Set<String> discoverMembers() {
LOGGER.trace("CELLAR EC2: discover members");
Set<String> members = new HashSet<String>();
DescribeInstancesResult describeInstancesResult =
amazonEC2.describeInstances(describeInstancesRequest);
List<Reservation> reservations =
describeInstancesResult.getReservations();
for (Reservation reservation : reservations) {
List<Instance> instances = reservation.getInstances();
for (Instance instance : instances) {
String privateIpAddress =
instance.getPrivateIpAddress();
if (privateIpAddress != null) {
LOGGER.debug("Adding EC2 instance: {} [{}]",
instance.getInstanceId(), privateIpAddress);
members.add(privateIpAddress);
}
}
}
return members;
}
I look forward to your advice, if you have any other questions feel free to ask.
Thanks,
Dario
-----Original Message-----
From: Jean-Baptiste Onofré [mailto:[email protected]]
Sent: 05 December 2015 05:58
To: [email protected]
Subject: Re: interested in a Amazon EC2 discovery service for Cellar?
Hi Dario,
Thanks for sharing !
I'm curious: do you use the cellar-cloud discovery ?
I'm interested about the step you did. Maybe you can write a blog post about
that (if you don't have a blog, I would be more than happy to post on mine) ?
Thanks again !
Regards
JB
On 12/04/2015 05:56 PM, Dario Cimafonte wrote:
> Hi,
>
> I did a copy-paste-amend exercise and in 2 hours I got a prototype
> cellar-ec2 discovery service working.
>
> Before I go through the effort of Jira, fork, pull request etc. I
> wanted to check whether anyone is interested in that.
>
> Thanks,
>
> Dario
>
--
Jean-Baptiste Onofré
[email protected]<mailto:[email protected]>
http://blog.nanthrax.net
Talend - http://www.talend.com
