|
Home > Archive > WebSphere Application Server > October 2006 > Continuous availability during EAR deployment (Websphere 6)
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Continuous availability during EAR deployment (Websphere 6)
|
|
|
| I am using Websphere ND 6.0.2.13 on a Unix platform.
I have a single cell, single cluster environment. My cell has two nodes.
I want to be able to re-deploy EAR's to the cell without interrupting
users who are logged into the application.
The IBM WAS6 Redbook suggests that this is possible:
"With Network Deployment, you can use application server clustering to enhance
workload distribution. A cluster is a logical collection of application server
processes that provides workload balancing and high availability.
Application servers in a cluster are members of that cluster and must all have
identical application components on them. Other than the applications on them,
cluster members do not have to share any other configuration data.
For example, one cluster member might run on a large multi-processor server
while another member of that same cluster might run on a small mobile
computer. The server configuration settings for each of these two cluster
members is very different, except the application components that are assigned
to them. In that area of configuration, they are identical.
The members of a cluster can be located on a single node (vertical cluster),
across multiple nodes (horizontal cluster), or on a combination of the two.
When you install, update, or delete an application, the updates are automatically distributed to all members in the cluster. In WebSphere Application Server V5, if you updated an application on a cluster, you had to stop the application on every server in
the cluster, install the update, and then restart the server. With WebSphere Application Server V6, the Rollout Update option allows you to update and restart the application servers on each node, one node at a time. This provides continuous availability
of the application."
Is anybody doing this? How does Websphere ND handle in-flight requests?
Sean
| |
| Paul Ilechko 2006-10-25, 1:36 am |
| sullivan.sean@menlolog.com wrote:
> I am using Websphere ND 6.0.2.13 on a Unix platform.
>
> I have a single cell, single cluster environment. My cell has two
> nodes. I want to be able to re-deploy EAR's to the cell without
> interrupting users who are logged into the application.
>
> The IBM WAS6 Redbook suggests that this is possible:
>
>
> "With Network Deployment, you can use application server clustering
> to enhance workload distribution. A cluster is a logical collection
> of application server processes that provides workload balancing and
> high availability. Application servers in a cluster are members of
> that cluster and must all have identical application components on
> them. Other than the applications on them, cluster members do not
> have to share any other configuration data. For example, one cluster
> member might run on a large multi-processor server while another
> member of that same cluster might run on a small mobile computer. The
> server configuration settings for each of these two cluster members
> is very different, except the application components that are
> assigned to them. In that area of configuration, they are identical.
> The members of a cluster can be located on a single node (vertical
> cluster), across multiple nodes (horizontal cluster), or on a
> combination of the two. When you install, update, or delete an
> application, the updates are automatically distributed to all members
> in the cluster. In WebSphere Application Server V5, if you updated an
> application on a cluster, you had to stop the application on every
> server in the cluster, install the update, and then restart the
> server. With WebSphere Application Server V6, the Rollout Update
> option allows you to update and restart the application servers on
> each node, one node at a time. This provides continuous availability
> of the application."
>
>
> Is anybody doing this? How does Websphere ND handle in-flight
> requests?
>
There are better ways to get continuous availability. I suggest you read
Pete Van Sickel's paper on this topic on developerworks. It's written
for WAS 5, but the concepts still apply. You typically don't want to
have different versions of the application running on different nodes at
the same time - it's very uncontrolled, and if something were to go
wrong it would be harder to recover. Having a backup cell or cluster and
using plugin configuration to switch is a more controlled approach.
| |
| Randy Schnier 2006-10-25, 1:36 am |
| In-flight requests are handled through a function called "server
quiesce." Basically, when each app server is told to shut down, it
actually goes into Quiesce mode. Existing requests are allowed to
complete (and requests coming in under an existing transaction are
allowed in) but new requests are directed to a different server within
the cluster. After a period of time in Quiesce (configurable), the
server actually shuts down and the app upgrade can then take place.
Transactions still in progress at the end of the quiesce period are
rolled back and the client receives a corresponding exception, giving it
the option to retry. However, the quiesce period is typically set so
that if a transaction has not completed by the end of the quiesce, it
would have exceeded the transaction timeout value anyway (and gotten the
same exception).
sullivan.sean@menlolog.com wrote:
>
With WebSphere Application Server V6, the Rollout Update option allows
you to update and restart the application servers on each node, one node
at a time. This provides continuous availability of the application.
> How does Websphere ND handle in-flight requests?
>
> Sean
>
|
|
|
|
|