r3 - 2018-06-11 - 14:50:25 - AlexBernsteinYou are here: TWiki >  Deployment Web > DeploymentInstallingUpgradingAndMigrating > UpgradingAClusterOfChangeAndConfigurationManagementApplications

Upgrading a cluster of Change and Configuration Management applications

Authors: MichaelAfshar, AlexBernstein, ChrisAustin, PrabhatGupta, YanpingChen, BreunReed
Build basis: Change and Configuration Management 6.0.5 and later

Upgrading clustered application requires very little custom steps. Essentially, it boils down to unclustering the application (see below), performing an upgrade, replicating the settings from the upgraded node and then re-enabling the cluster.

Upgrade Steps

  • Stop all Liberty profiles in order (applications first, then JTS and Distributed Cache Microservice (DCM) last).
  • Leave the load-balancer (HAProxy) running and unmodified (unless changes are specifically needed for the new version).
  • Remove the com.ibm.team.repository.mqtt.broker.address line from teamserver.properties of any clustered apps (This is optional, only do this if repotools -addTables fails to execute).
  • Upgrade CLM normally (JTS first, microservice and apps last).
    • Install the new release (using Installation Manager, or unzipping the build) on all servers, including all nodes of a cluster.
    • Manually review/update to ensure any Liberty JVM options and keystores are preserved.
    • For clustered applications, run the upgrade script(s) on only one node of each cluster (though if you accidentally upgrade on both nodes, nothing will break - the second "upgrade" won't actually change anything on the database).
  • Re-add the original com.ibm.team.repository.mqtt.broker.address line to teamserver.properties (if removed).
  • On any additional clustered nodes, copy the teamserver.properties from the upgraded node.
  • Make sure the new server startup scripts of clustered applications have cluster-related entries, such as unique cluster Node IDs.
  • Manually review/update distributedCache.cfg to preserve any customizations as well as new values needed by the new version. Copy the keystore.
  • Start all profiles in order (DCM and JTS first; verify they are working, then start the other apps).
  • Run all diagnostics and test all applications.

Notes on Distributed Cache Microservice (DCM)

Newer versions of DCM monitor the online status of the applications' nodes. When a server on a node machine is shut down, the signal is sent to DCM. After all nodes of a cluster went offline gracefully, the DCM clears out all persisted distributed data for that cluster. The cluster specific storage is just a sub-folder under <DCM-INSTALL-LOCATION>/clustering/cache/distributedData folder. The name of that sub-folder is the same as the name of the cluster. This is implemented so that when applications' nodes are coming up after planned downtime, they initialized their distributed data from scratch instead of using obsolete and potentially stale persisted copy. The persisted copy is used only be the node that is coming up after a crash or another unplanned outage and needs the current copy (a snapshot) of the distributed data.

While performing the upgrade it is always a good idea to ensure that while everything is shut down, no persisted data is left behind. If there is anything in the folder that represents your cluster, you can safely delete its contents. Only do this while the application server is offline.

This is little trickier when the same DCM is used centrally for several clusters. Bringing down and upgrading "shared" DCM may affected applications that are not to be upgraded.

Never use DCM from the prior release with the newer CLM. Always copy customized settings from an existing DCM into the one that came with the newer CLM. Run DCM distributed with the new CLM.

NOTE: Prior to CLM version 6.0.5 there was no Distributed Cache Microservice; its functionality was part of JTS. If upgrading from 6.0.4 simple follow instructions for installing and registering fresh DCM from scratch.


Related topics: Deployment web home, Mode details about clustered configurations

External links:

Additional contributors: TWikiUser, TWikiUser

Edit | Attach | Printable | Raw View | Backlinks: Web, All Webs | History: r5 < r4 < r3 < r2 < r1 | More topic actions...
This site is powered by the TWiki collaboration platformCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Contributions are governed by our Terms of Use. Please read the following disclaimer.
Ideas, requests, problems regarding the Deployment wiki? Create a new task in the RTC Deployment wiki project