10:31 a.m.
Hi All:
I'd like to hear any and all clustering use cases relevant to tooling
that any of you have run into. Whether it be creating multiple servers /
runtimes / profiles, actions you'd like supported on a cluster type node
in the view, anything you can imagine. Spit it out. Speak up. Or
forever hold your peace ;)
4:58 p.m.
I run into three big (Fortune 50) clients that want to cluster across data centers. Not
saying it is a good idea, just saying.
Robb Greathouse
JBoss
CELL 505-480-4639
OFFICE 505-255-2652
FAX 505-554-2834
----- Original Message -----
From: "Rob Stryker" <rob.stryker(a)redhat.com>
To: jbosstools-dev(a)lists.jboss.org
Sent: Friday, April 3, 2009 2:31:16 AM GMT -06:00 Central America
Subject: [jbosstools-dev] Cluster Feature Collecting
Hi All:
I'd like to hear any and all clustering use cases relevant to tooling
that any of you have run into. Whether it be creating multiple servers /
runtimes / profiles, actions you'd like supported on a cluster type node
in the view, anything you can imagine. Spit it out. Speak up. Or
forever hold your peace ;)
_______________________________________________
jbosstools-dev mailing list
jbosstools-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/jbosstools-dev
1:26 a.m.
Robb Greathouse wrote:
I run into three big (Fortune 50) clients that want to cluster across
data centers. Not saying it is a good idea, just saying.
To be clear, I'm talking only about tooling here. I'm not thinking about
pushing use cases to the runtime or anything like that, but simply
asking tool users how they currently use clustering, and how they'd like
cluster-interaction inside eclipse to function.
What deployment options should be supported? If I set up 5 local AS
instances, for example, do I need to deploy to all 5 or just one? Do
they copy to each other?
At development time, are most clusters on the same machine? Are they
remote? Are they local? Do you develop on a local cluster and then just
deploy it to the remote one later? For remote ones, how do you deploy?
FTP? SCP? Should remote deployment options be included? Should remote
deployment occur after every change, or only on command? Do you deploy
to one remote machine or all 5 nodes?
If you only need to deploy to one server, then the view can include just
one "wrapper" type server, whereas if you have to deploy to each of them
separately, they're probably better off modeled as individual servers
then grouped under a cluster node.
Robb Greathouse
JBoss
CELL 505-480-4639
OFFICE 505-255-2652
FAX 505-554-2834
----- Original Message -----
From: "Rob Stryker" <rob.stryker(a)redhat.com>
To: jbosstools-dev(a)lists.jboss.org
Sent: Friday, April 3, 2009 2:31:16 AM GMT -06:00 Central America
Subject: [jbosstools-dev] Cluster Feature Collecting
Hi All:
I'd like to hear any and all clustering use cases relevant to tooling
that any of you have run into. Whether it be creating multiple servers /
runtimes / profiles, actions you'd like supported on a cluster type node
in the view, anything you can imagine. Spit it out. Speak up. Or
forever hold your peace ;)
_______________________________________________
jbosstools-dev mailing list
jbosstools-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/jbosstools-dev
10:19 a.m.
If you only need to deploy to one server, then the view can include
just one "wrapper" type server, whereas if you have to deploy to each
of them separately, they're probably better off modeled as individual
servers then grouped under a cluster node.
I would say clustering might be two or
three things:
1) Being able to group arbitrary group of (jboss) servers on which you
can do the various operations (i.e. stop, start, publish, etc.).
That grouping possibly have to be sequenced/ordered, but i'm not
sure how necessary that is.
2) Easy creation of a server that have the same setup as anyone else in
the group - except for a few key settings (I think Galder got those in a
jira somewhere?)
This might require physically creating/copying a server
configuration on the disk but that would have to be optional in case of
local vs remote servers.
This feature might not just be for cluster usecases but also for
normal server creation.
p.s. how about raising this on the forums where users are more than here ?
/max
8:52 p.m.
Hey Rob,
Sorry I never replied to this email. I've been meaning to do so but keep
having to do other things. So, let me explain all the steps I have to do
to create a cluster and have it ready to run a test:
Command line:
- cd AS_HOME/EAP_HOME (assume: EAP 4.3.0.GA)
- cp -r server/all server/xyz1
- cp -r server/all server/xyz2
Open JBDS:
- Create a new runtime and server for xyz1, give it a name:
eap-4.3.0.ga-xyz1 and add bind address localhost2
- Create a new runtime and server for xyz2 and give it a name:
eap-4.3.0.ga-xyz2, add bind address localhost3 and give it a different
JBoss Messaging server Id.
- Deploy your apps in xyz1 and xyz3
As you can see, if you wanna create a simple cluster for testing, it
takes a fair few clicks and operations to get it done.
I envision eventually having some wizard or GUI window from where you
could basically do all this in one go:
- Select EAP/AS home.
- Select number of configs to cluster, i.e. 2
- Give names to these X number of configs:
- Give bind addresses (or select service binding port configs) for each
config.
- Optionally give some special system properties to each, i.e. JBoss
Messaging server id.
- What deployments go in each server.
Underneath, this would do what I know have to do manually.
Thoughts?
Rob Stryker wrote:
Hi All:
I'd like to hear any and all clustering use cases relevant to tooling
that any of you have run into. Whether it be creating multiple servers /
runtimes / profiles, actions you'd like supported on a cluster type node
in the view, anything you can imagine. Spit it out. Speak up. Or
forever hold your peace ;)
--
Galder Zamarreño
Sr. Software Maintenance Engineer
JBoss, a division of Red Hat
10:17 a.m.
Command line:
- cd AS_HOME/EAP_HOME (assume: EAP 4.3.0.GA)
- cp -r server/all server/xyz1
- cp -r server/all server/xyz2
This feature is now in nightly build afaik
see https://jira.jboss.org/jira/browse/JBIDE-3896
Feedback welcome.
Open JBDS:
- Create a new runtime and server for xyz1, give it a name:
eap-4.3.0.ga-xyz1 and add bind address localhost2
- Create a new runtime and server for xyz2 and give it a name:
eap-4.3.0.ga-xyz2, add bind address localhost3 and give it a different
JBoss Messaging server Id.
I assume you have setup localhost2, localhost3, ... in your OS hosts setup ?
- Deploy your apps in xyz1 and xyz3
As you can see, if you wanna create a simple cluster for testing, it
takes a fair few clicks and operations to get it done.
I envision eventually having some wizard or GUI window from where you
could basically do all this in one go:
- Select EAP/AS home.
- Select number of configs to cluster, i.e. 2
- Give names to these X number of configs:
- Give bind addresses (or select service binding port configs) for
each config.
The lower parts support all of the above now (just no unified ui).
- Optionally give some special system properties to each, i.e. JBoss
Messaging server id.
examples ?
- What deployments go in each server.
Do you mean the
deployments will be different ?
Underneath, this would do what I know have to do manually.
Thoughts?
I actually think we should add the core level flexibility for clustering
before doing the full wizard. Not sure though.
What i'm seeing missing is a way to "group" servers in the view and for
operations.
i.e. I think it would be interesting to see if we can utilize the
"working set" concept that exists for resources in eclipse to servers.
Such a "Server working set" is a way to "tag" servers (multiple
servers
can be in multiple groups) and it is something that you
can do operations on. i.e. "Start" on a working set will start all
servers in there, "Deploy" on a working set will deploy the chosen
project to all servers
in that working set. etc.
WDYT ?
-max
3:03 p.m.
Max Rydahl Andersen wrote:
>
> Command line:
> - cd AS_HOME/EAP_HOME (assume: EAP 4.3.0.GA)
> - cp -r server/all server/xyz1
> - cp -r server/all server/xyz2
This feature is now in nightly build afaik
see https://jira.jboss.org/jira/browse/JBIDE-3896
Feedback welcome.
>
> Open JBDS:
> - Create a new runtime and server for xyz1, give it a name:
> eap-4.3.0.ga-xyz1 and add bind address localhost2
> - Create a new runtime and server for xyz2 and give it a name:
> eap-4.3.0.ga-xyz2, add bind address localhost3 and give it a different
> JBoss Messaging server Id.
I assume you have setup localhost2, localhost3, ... in your OS hosts
setup ?
> - Deploy your apps in xyz1 and xyz3
>
> As you can see, if you wanna create a simple cluster for testing, it
> takes a fair few clicks and operations to get it done.
>
> I envision eventually having some wizard or GUI window from where you
> could basically do all this in one go:
>
> - Select EAP/AS home.
> - Select number of configs to cluster, i.e. 2
> - Give names to these X number of configs:
> - Give bind addresses (or select service binding port configs) for
> each config.
The lower parts support all of the above now (just no unified ui).
> - Optionally give some special system properties to each, i.e. JBoss
> Messaging server id.
examples ?
JBoss Messaging server Id is an example :)
> - What deployments go in each server.
Do you mean the deployments will be different ?
You could be deploying the same thing accross the cluster, or u could be
deploying ear1 in node1 and ear2 in node2.
>
> Underneath, this would do what I know have to do manually.
>
> Thoughts?
I actually think we should add the core level flexibility for clustering
before doing the full wizard. Not sure though.
What i'm seeing missing is a way to "group" servers in the view and for
operations.
i.e. I think it would be interesting to see if we can utilize the
"working set" concept that exists for resources in eclipse to servers.
Such a "Server working set" is a way to "tag" servers (multiple
servers
can be in multiple groups) and it is something that you
can do operations on. i.e. "Start" on a working set will start all
servers in there, "Deploy" on a working set will deploy the chosen
project to all servers
in that working set. etc.
WDYT ?
+1, sounds about right. A way to get those members united in something
that is meaningful at the Eclipse level would be handy.
-max
--
Galder Zamarreño
Sr. Software Engineer
Infinispan, JBoss Cache
5413
days inactive
5495
days old
jbosstools-dev@lists.jboss.org
6 comments
4 participants
participants (4)
-
Galder Zamarreno -
Max Rydahl Andersen -
Rob Stryker -
Robb Greathouse