Remember that only global administrators and promotion administrators can deploy revisions. More information is available in the Authorization section.
A deployment process consists in executing all the changes defined in one or more revisions in the Denodo servers of an environment. More specifically:
- If a revision has Virtual DataPort elements, the corresponding VQL will be executed in all the Virtual DataPort servers that belong to the environment.
- If a revision has Scheduler tasks, they will be created (for CREATE revisions) or removed (for DROP revisions) from the first Scheduler server of each cluster that belongs to the environment.
- If the environment has Data Catalog server synchronization configured, the corresponding Data Catalog servers metadata will be synchronized according to the provided configuration.
The Solution Manager executes first all the revisions in one server before continuing to the next one. The order in which the servers are selected depends on how they are arranged in the catalog. Therefore, the first server of the first cluster of the target environment will be the first server in updating its metadata.
Those clusters with no Virtual DataPort servers will be ignored during the deployment process.
The Scheduler tasks are only created / removed and executed (if needed) in the first Scheduler server of each cluster of the environment.
The next picture summarizes the basic steps of a generic deployment process:
The deployment configuration of the target environment defines which actions are actually performed during the deployment process.
The Solution Manager performs some basic validations before starting the execution of a deployment in an environment:
- Deployments must be enabled in the configuration of the target environment.
- The target environment must define all the Virtual DataPort properties that parameterize the VQL of the revision.
- Every cluster of the target environment with Scheduler servers must define all Scheduler properties that parameterize the tasks of the revision.
- There is no other deployment in progress in the target environment.
- The environment configuration matches the current environment topology.
Additionally, the Solution Manager performs another validations depending on the deployment configuration of the environment. These validations are explained in the following section.
The Solution Manager offers several options for performing the deployment process. This section evaluates the advantages and trade-offs of each one of them.
This option assumes that the application service will not be available during the whole deployment process. It is intended for offline promotion scenarios.
The Solution Manager deploys the revisions in all the corresponding servers, without explicitly disabling any cluster or server in the load balancer.
There is no need to perform any additional validation in simple deployments.
Deployment Without Service Interruption¶
Intended for online promotion scenarios, this option maintains the application service available during the deployment process. However, there may be inconsistencies while the deployment lasts, since servers with old and new versions of the metadata might provide service at the same time.
To maintain the application service, the Solution Manager executes scripts to disable Virtual DataPort servers or clusters in the load balancer and to enable them again when the deployment finishes. There are two promotion strategies available that control the interaction with the load balancer:
- One by one: The promotion is executed in an incremental way. Only one server or cluster is disabled (and then enabled) at a time. This results in longer times of inconsistency. The inconsistency time grows with the number of clusters or servers in the target environment.
- Half by half: In this strategy half of the servers or clusters are disabled (and then enabled) at the same time. This results in inconsistency for a short time.
Take into account that the promotion strategy does not affect the deployment process duration. No matter how many servers or clusters are disabled in the load balancer at a time, the final servers will be updated one by one.
The Solution Manager only enables servers whose deployment process finished successfully. In the same way, the Solution Manager only enables those clusters where the deployment process was successful for all their servers.
The options available for the deployments without service interruption depend on the topology of the target environment.
Active Passive Load Balancing Considerations¶
An active/passive cluster consists of one or more active servers and one or more passive standby servers with a load-balancer to determine when failover is required.
In an environment with this architecture we recommend to create / order the servers inside of each cluster putting the passive servers before the active servers.
Deployment Without Service Interruption in a Single Cluster Environment¶
This scenario is designed for environments with only one cluster with at least one Virtual DataPort server. In this case, the Solution Manager Tool allows to configure:
- The server promotion strategy. This policy establishes if the Virtual DataPort servers are disabled in the load balancer one by one or half by half.
- If a secondary shared cache is available. If enabled, it is necessary to provide a main and a secondary cache URL.
Secondary shared cache option is useful in those scenarios with shared cache between all servers. During a deployment with this option enabled, the Solution Manager:
- Calculates the new cache URL. First, it detects which one of the two shared caches (the primary or the secondary) is configured in the first Virtual DataPort server. The other cache, the one that is not currently used, will be selected as the new shared cache.
- Swaps the cache for each Virtual DataPort server to the new cache before deploying all the revisions in that server.
- Executes all Scheduler VDP Cache tasks in order to populate the cache, after deploying the revisions in the first Virtual DataPort server.
Secondary shared cache option only applies to those deployments with at least one CREATE revision that contains a Scheduler VDP Cache task.
Currently, the cache swap process is limited to global cache, this is, there are no caches defined for particular databases.
In this scenario, the Solution Manager performs the following additional validations:
- The target environment must provide scripts to disable and enable Virtual DataPort servers in the load balancer.
- All load balancing variables that participate as arguments of the scripts must have a value defined for all the Virtual DataPort servers and clusters that belong to the target environment of the deployment.
- All Virtual DataPort servers must have the cache enabled, if secondary shared cache is enabled and at least one CREATE revision contains a Scheduler VDP Cache task.
Deployment Without Service Interruption in a Several Clusters Environment¶
This scenario is designed for environments with more than one cluster with at least one Virtual DataPort server. In this case, the Solution Manager Tool allows to configure the cluster promotion strategy. This policy establishes if the clusters with Virtual DataPort servers are disabled in the load balancer one by one or half by half.
In this scenario, the Solution Manager performs the following additional validations:
- The target environment must provide scripts to disable and enable clusters in the load balancer.
- All load balancing variables that participate as arguments of the scripts must have a value defined for all the clusters that belong to the target environment of the deployment.
When a script is executed, it fails if the execution time reaches the timeout
defined in the property
of the file
This timeout is measured in seconds. Increase this timeout if the
scripts perform complex operations that need more time
Solution Manager has the following limitations related with cache:
- It is not possible to define Virtual DataPort properties at cluster level. In the case of cache data sources in multicluster environments, it is necessary to configure the cache manually. Due to this limitation, you should not include any cache datasources in CREATE revisions, otherwise during the deployment process the cache configuration of all clusters will be overwritten, leaving one of them in a possible inconsistent status. Solution Manager shows a warning message when deploying a revision with the features above:
- The cache swap process (in deployments in environments with a single cluster and cache swap enabled) only works with global cache (it does not work with caches defined for particular databases).