Application Deployment

Follow these guidelines for application deployment.

When describing managing the installation and un-installation of a Grid application on a host, we use the terms deploy and undeploy. The Grid contains an application deployment mechanism which supports managing applications deployment on one or more hosts including the primary install (deploying to the first host), scaling-in, scaling-out, upgrading and un-installing. To be installed a grid application must be deployed on at least one host, otherwise the application is considered as un-installed.

An application's deploy state indicates, per host or globally, the current deploy state of the application - that is, if the deploy was successful, is in progress or failed.

A deploy is considered successful if the application files could be unzipped on disk on the target host and the deployment profile was executed successfully.

Note that an application binding can only be started on a host where the application is considered to be OK. An application that is currently being deployed can have a number of transitive states (<op> is Deploy, Undeploy, or Run Profile):

<op> Pending The operation has been scheduled to run.
<op> Launching The Agent has issued the node start command for the deploy.
<op> Launched The deploy node launched successfully.
Pre-<op> in Progress The preDeploy or preUndeploy method of the application's deployment profile is being executed (only applicable for Upgrade and Undeploy).
<op> in Progress The application files are being unzipped (or removed for Undeploy) on the target host(s).
Post-<op> in Progress The postDeploy method of the application's deployment profile is being executed (not applicable for Undeploy).

If the deploy state is "<op> in Progress (held)" it was launched in debug mode and must be manually released by clicking the deploy state link and then the Release button.

Launching a deploy operation in debug is only available when the Show Developer Features option is enabled in the UI.

If the deploy state is "<op> ... (retrying)" the operation failed for some reason and will be retried until successful.

The most common reason is that a file could not be written (or removed) in the file system - check the deploy log for more details.

Note that if the deploy node should crash or is killed, the deploy will be considered to have failed.

If a deploy fails for any reason, one of the following states is reported:

Deploy Failed
Upgrade Failed Run Profile Failed
Undeploy Failed

If any fail state is appended with "Partially" it means that the application is OK on at least one host (and thus operational), but failed on one or more hosts.

For any of the failed states there is always an option to retry or cancel the failed operation on the application's page in the Config Manager.

It is also possible to view the log file for the failed deploy which will indicate exactly what the failure was and when and where (on which host) the problem was reported.

Also note that deploy nodes will always run on the primary host only (the host where the Registry is located) regardless of the target host(s) for the deploy operation.