Caché Installation Guide
Upgrading Caché
[Back] [Next]
   
Server:docs1
Instance:LATEST
User:UnknownUser
 
-
Go to:
Search:    

This chapter is intended for customers who are upgrading from Caché release 4.1 and later. Topics in this chapter include:

Important:
Before upgrading, review the following documents:
Caution:
Prior to beginning an upgrade of a Caché or Ensemble instance, it is essential that the instance be shut down cleanly. To verify that the shutdown was clean, examine the cconsole.log file after the shutdown finishes. If the log contains entries similar to the following, then the shutdown was clean:
...
05/03/11-14:24:13:234 (5204) 0 Journal restore not required at next startup
05/03/11-14:24:13:234 (5204) 0 Transaction rollback not required at next startup
...
If these entries are not present, the instance did not shut down cleanly. Please contact the InterSystems Worldwide Response Center before proceeding with the upgrade.
Note:
If you run Caché MultiValue, some files may be overwritten during upgrade; see the section “Notes on Files Overwritten During Upgrade” in the chapter Installing Caché in Using the MultiValue Features of Caché for more information.
Supported Upgrade Paths
You can upgrade directly to this release of Caché from release 2009.1 or any later release. However, if you are upgrading a pre-2012.1 instance containing any 2K block size databases, these must be converted to 8K format before upgrading to version 2012.x or beyond. In this situation, use the following procedure:
  1. Upgrade the instance to version 2011.1, if necessary.
  2. Use the SYS.Database.Copy() method to convert all 2K databases to 8K format.
  3. Upgrade the instance from 2011.1 to the current release.
See Configuring Databases in the “Configuring Caché” chapter of the Caché System Administration Guide for information about database block size.
Note:
To upgrade from a version older than 2009.1, you must first upgrade to 2009.1 or later, then upgrade to the current release. Use the procedure for converting 2K block size databases if applicable. For assistance with upgrading from earlier versions, contact the InterSystems Worldwide Response Center (WRC).
Upgrading to a Maintenance Release
Most of the information in this chapter applies to upgrading Caché to a new major version. If you are upgrading to a maintenance release of the installed version of Caché, review the following:
Upgrade Tasks
The following upgrade tasks are necessary regardless of the platform on which you are upgrading and running Caché. Perform these tasks before you run the Caché installation procedures:
  1. Obtain an updated license key — the key structure changed in Caché 5.1; upgrades from Caché 5.0 or earlier require an updated key.
  2. Back up the system — before upgrading, InterSystems recommends that you run a complete backup of your system. Use your customary full operating system backup procedures.
  3. Check system integrity — run a system integrity check on existing directories to ensure there is no corruption in any of the databases.
  4. Save custom classes, routines and globals — to prevent your own classes, routines and globals in the %SYS namespace from being affected by the upgrade installation, ensure that they have names that begin with “Z”, “z”, “%Z”, or “%z”. All .int and .obj routines (except for Z*, z*, %Z*, and %z*) are deleted from the %SYS namespace when upgrading.
    On an upgrade, the CACHELIB, CACHETEMP, DOCBOOK, CACHE, and SAMPLES databases are completely replaced.
    Any .mac or .inc routines are not affected during the upgrade.
  5. Save user files — to ensure that your user files are not deleted or replaced during upgrades, save them in the install-dir\Mgr\User directory, which is the only directory that is not subject to modification during upgrades (that is, except for the install-dir\Mgr\User directory, any files and directories in the Caché install directory may be deleted or replaced during upgrades).
  6. Review encryption settings — because of a limitation in prior releases, if you are upgrading from release 2010.1 or earlier and have encryption set to interactively ask for the username, password, and key file, you should temporarily change this setting to load encryption information automatically before upgrading (see Configuring Startup with Unattended Key Activation in the “Managed Key Encryption” chapter of the Caché Security Administration Guide for more information). After you perform the upgrade, you should then reenable the interactive setting. (Failure to configure startup with unattended key activation may cause the upgrade to fail; if this happens, start the system normally, change the startup configuration to unattended mode, and then reinstall the new version.)
    If you are upgrading using an .rpm or .dmg file, you must disable database encryption before upgrading and reenable it when the upgrade is complete.
In addition, there are other tasks that may be necessary depending on your environment and the components as described the following sections:
Upgrading a Cluster
This section describes general considerations to take into account when upgrading Caché in a Caché failover cluster (as described in the failover cluster appendixes, such as Using Windows Clusters with Caché, of the Caché High Availability Guide) or in a Caché concurrent cluster (as described in the Caché Concurrent Clusters chapter of the Caché System Administration Guide).
Note:
The information in this section does not apply to systems in which ECP application servers are deployed. For information about upgrading ECP configurations, see Upgrading ECP Configurations.
Important:
At a general level, it is important to:
In Windows, for example, to prevent cluster failover/automatic Caché startup, always tale the Caché cluster resource offline prior to the upgrade; when ready to resume operation, bring the resource back online. Similar steps should be followed on all other platforms; refer to the failover software product documentation that is relevant to your installation.
For platform-specific information, see the following:
Upgrading Caché on Windows Failover Clusters
On Microsoft Windows clusters, use the following procedure to upgrade Caché.
Important:
Do not start and stop Caché from the Caché Launcher. Instead, using Failover Cluster Management, take the Caché cluster resource offline to stop Caché, and bring the Caché cluster resource online to start Caché.
It is important to upgrade Caché on both systems in a Windows cluster before resuming cluster operation to ensure that the correct registry settings are maintained. For this reason, you cannot upgrade a Windows cluster without making the cluster unavailable for a period of time.
  1. After ensuring that a cluster failover will not occur automatically, shut down on the currently running (primary) system (for example, System A) by taking the Caché cluster resource offline. This step is recommended to quiesce the system prior to an upgrade.
  2. Upgrade Caché on System A. While the upgrade process will start Caché on System A after the upgrade is successful, it is recommended that users (direct-connect, background processes, ECP application servers, etc.) not be allowed back on to the system until the entire upgrade process (cluster-wide) is complete.
  3. Shut down Caché on System A by taking the Caché cluster resource offline.
  4. Fail the cluster over to the secondary system (for example, System B).
  5. Upgrade Caché on System B. While the upgrade process will start Caché on System B after the upgrade is successful, it is recommended that users (direct-connect, background processes, ECP application servers, etc.) not be allowed back on to the system until the entire upgrade process (cluster-wide) is complete.
  6. Fail the cluster over to System A.
  7. Start Caché on System A by bringing the Caché cluster resource online.
  8. Ensure that the cluster resource type (for example, Caché ISCCres2003) is brought online so that automatic failover can occur. Users (direct-connect, background processes, ECP application servers, etc.) can be permitted to reconnect to Caché (on System A) at this point.
Upgrading Caché on UNIX®/Linux Failover Clusters
On UNIX/Linux systems clusters, use the following procedure to upgrade Caché:
  1. Gather the list of Caché instances on the two failover nodes. This can be done by using the ccontrol list command. Note the list of systems (and paths, etc.), if any, that are unique to the second (standby) failover node.
  2. After ensuring that a cluster failover will not occur automatically, shut down Caché on the currently running (primary) system (for example, System A).
  3. Upgrade Caché on System A. The upgrade process should start Caché once the upgrade is complete. Users (direct-connect, background processes, ECP application servers, etc.) can be permitted to reconnect to Caché (on System A) at this point.
  4. Copy the /usr/local/etc/cachesys/ directory from System A to System B. Note: please ensure that the permissions and ownership for the cachesys directory as well as the files within that directory remain intact.
  5. If there are any Caché instances that were unique to System B, you must recreate that information manually using the ccontrol create command to add the instance information to the registry. The correct usage of the command can be found by typing ccontrol create help at the OS prompt.
  6. Ensure that the appropriate cluster resource associated with Caché is brought back online so that automatic failover can occur.
Upgrading Caché on OpenVMS Concurrent Clusters
Prior to version 2009.1, Caché on OpenVMS Clusters required a “cold shutdown” when upgrading Caché. However, beginning with Caché version 2009.1, when upgrading to a newer version, you can perform a “rolling upgrade” without having to delete the PIJ file.
Upgrading 2008.2 and Earlier Versions of Caché
On OpenVMS clusters, use the following procedure to upgrade Caché (version 2008.2 and earlier):
  1. Perform a “cold shutdown” — Shut down all members of the cluster in preparation of the upgrade.
  2. Remove the Pre-Image Journal (PIJ) file — If you are upgrading a member of a Caché concurrent cluster, cleanly shut down all members of the cluster and remove the CACHE.PIJ file.
    Note:
    If you do not remove the CACHE.PIJ file, the installation is not upgraded and error messages, such as the following, are written in the cconsole.log for startup:
    Cache (2100036c) Tue Aug 1 14:28:59 2007
    Activating Namespaces
    Cache (21000404) Tue Aug 1 14:28:59 2007 Cluster image journal is 
    incompatible with this version 
    Cache (21000404) Tue Aug 1 14:28:59 2007 Unable to join the cluster 
    Cache (21000404) Tue Aug 1 14:29:00 2007
    ENQdaemon exited due to VMS error code (decimal) 0
  3. Perform the upgrade on each node. As each node is upgraded, Caché on that node starts up once the upgrade is complete. It is recommended that users be prevented from connecting to the cluster till all nodes have been upgraded.
Upgrading 2009.1 and Later Versions of Caché
On OpenVMS clusters, use the following procedure to upgrade Caché (version 2009.1 and later):
  1. Leave all the nodes running.
  2. Perform upgrades on each node, one at a time.
    Note:
    When a node is upgraded, Caché is shut down momentarily and restarted upon completion of the upgrade.
Minimum Downtime Upgrade with Mirroring
This section provides instructions for upgrading Caché or Ensemble instances, an application, or both on the members of a Caché mirror.
As noted in the Mirroring chapter of the Caché High Availability Guide, all failover and DR async members of a mirror must be of the same Caché version, and can differ only for the duration of an upgrade. Once an upgraded member becomes primary, you cannot make use of the other failover member and any DR async members (and in particular cannot allow them to become the primary) until the upgrade is completed.
Mirroring does not require reporting async members to be of the same Caché version as the failover members, although application functionality may require it; see the “Supported Version Interoperability” chapter of InterSystems Supported Platforms for information about reporting async version requirements.
There are four mirror upgrade procedures, as follows. The first three procedures are designed to let you perform the upgrade you need with the shortest possible application downtime; they differ to accommodate different upgrade circumstances. The last, which is a bit simpler, can be used for any type of upgrade when minimizing downtime is not a priority.
Choosing a Mirror Upgrade Procedure provides an overview of the factors that determine which procedure you should use. Your circumstances may call for additional steps and details in a given procedure. If you are unsure which procedure is appropriate for you or whether a particular procedure is appropriate as stated, please contact the InterSystems Worldwide Response Center (WRC).
To avoid repetition, Minimum Downtime Upgrade Terms defines the terms that are used in multiple procedures.
Important:
On Linux systems supporting systemd, although it is possible to use either systemctl commands or the /etc/init.d/ISCAgent script to manage the ISCAgent, you must choose one method and use it exclusively; the ISCAgent should always be stopped using the method with which it was started (see Starting the ISCAgent on Linux Systems in the “Mirroring” chapter of the Caché High Availability Guide).
When you upgrade Caché on such a Linux system, a running ISCAgent is automatically restarted using systemd. If you are using the /etc/init.d/ISCAgent script to manage the ISCAgent, stop the agent before performing the upgrade so that it is not automatically restarted, then restart it using the script after the upgrade.
Choosing a Mirror Upgrade Procedure
There are two primary factors to consider in choosing the procedure appropriate for your mirror upgrade:
Whenever you upgrade to a maintenance release and are not making any application upgrades, you do not have to consider the second question; you can always use the simple Maintenance Release Upgrade procedure, which renders the mirror unavailable only for the time it takes to execute a planned failover.
However, when you upgrade from one major Caché version to another, perform application upgrades, or both, there is the possibility that your upgrade will involve changes to mirrored databases. Such changes must occur on the functioning primary failover member as part of the upgrade procedure, while application access is disabled. As a result, the upgrade requires more downtime than if you are not making any mirrored database changes. (The changes will then be replicated by the mirror on the backup and any async members.)
Following a major upgrade, InterSystems recommends that you recompile all classes in all application namespaces on the instance, including Ensemble namespaces, and some routines must also be recompiled (see Upgrade Specifics in the “General Upgrade Information” chapter of the Caché Recent Upgrade Checklist). Any application upgrade you perform may require changes to application data. In either of these situations, your upgrade may result in changes to mirrored databases. Therefore, if you are upgrading to a major release, are performing application upgrades, or both, you must determine whether any of the following conditions apply:
If your major upgrade and/or application upgrade involves any of these conditions, use the Major Upgrade (Mirrored Database Changes) procedure, which renders the mirror unavailable only for the time it takes to execute a planned failover and make the required mirrored database changes.
If your upgrade does not involve any of the listed mirrored database changes, consider the Major Upgrade (No Mirrored Database Changes) procedure, which renders the mirror unavailable only for the time it takes to execute a planned failover.
If you have a significant planned maintenance window and minimizing mirror downtime is not a major concern, you may choose to use the simpler Major Upgrade with Planned Downtime procedure instead.
In the case of major upgrades, the general sequence of each procedure applies to upgrades of the mirror’s Caché instances, application code, or both; adapt individual steps depending on what you are upgrading.
Adding Mirror Members During an Upgrade
If you plan to add members to a mirror, you may want to wait until a planned upgrade and add members of the new version, so that you will not have to upgrade them later. You can always add members of a newer version to a mirror during an upgrade, with one restriction: once a member of the newer version becomes primary, it must remain primary until all other members have been upgraded.
Adding new members during an upgrade can be very helpful when migrating a mirror to new hardware. Rather than upgrading one of the failover members before failing over to it, you can install a new instance of the new version on the target system, add it to the mirror as a DR async member, promote it to backup and then fail over to it, thus migrating the mirror primary to the new system. By repeating this technique you can then migrate the remaining failover member. The procedures in this section include steps for adding a new backup of the new version as an alternative to upgrading the existing backup, to illustrate this approach.
Minimum Downtime Upgrade Terms
In the procedures in this section, the following terms are used:
Maintenance Release Upgrade
If you are upgrading to a Caché maintenance release rather than a new major Caché version and are not making any application changes, use the following procedure, which renders the mirror unavailable only for the time required to execute a planned failover:
  1. To prevent failover from occurring until the backup is fully upgraded, set no failover.
  2. Perform one of the following two operations:
  3. Ensure that the backup (B or C) has become active by viewing the Mirror Monitor.
  4. Clear no failover.
  5. Perform a graceful shutdown of the primary (A). The mirror fails over and the backup (B or C) takes over as primary.
  6. Upgrade system A with the new version of Caché. System A becomes backup.
  7. If system C became primary and system B was demoted to DR async, upgrade system B.
Major Upgrade (Mirrored Database Changes)
If you are upgrading to a new major Caché version and/or performing an application upgrade and have determined that changes to mirrored databases are required (as described in Choosing a Mirror Upgrade Procedure), use the following procedure, which renders the mirror unavailable only for the time it takes to execute a planned failover and make the required mirrored database changes:
  1. To prevent failover from occurring until the backup is fully upgraded, set no failover.
  2. Perform one of the following two operations:
  3. Ensure that the backup (B or C) has become active by viewing the Mirror Monitor.
  4. Disallow new user access to the mirror using established practices at your site. Additionally, you must disable any application jobs that would normally start on system B when it becomes primary. If using Ensemble, disable production auto-start.
  5. Clear no failover.
  6. Perform a graceful shutdown of the primary (A). The mirror fails over and the backup (B or C) takes over as primary
  7. Upgrade mirrored classes and routines on the new primary (B or C). If an application upgrade requires further changes to mirrored databases, make these changes.
  8. Restore user access to the mirror. If using Ensemble, enable production auto-start and start productions.
  9. To prevent system A from taking over as primary until fully upgraded, set no failover
  10. Upgrade system A with the new version of Caché. System A becomes backup.
  11. Upgrade non-mirrored classes and routines on system A, if any.
  12. Clear no failover.
  13. If system C became primary and system B was demoted to DR async, upgrade system B and upgrade non-mirrored classes and routines on system B, if any.
Major Upgrade (No Mirrored Database Changes)
If you are upgrading to a new major Caché version and/or performing an application upgrade and have determined that changes to mirrored databases are not required (as described in Choosing a Mirror Upgrade Procedure), you may be able to use the following procedure.
This procedure is simplest for static applications where separate non-mirrored routine databases are always maintained. It can, however, be used even if the separate routine databases are normally mirrored by removing these databases from the mirror for the duration of the upgrade.
Important:
If the routine databases are normally mirrored, ensure that they do not contain any application data before removing them from the mirror.
This procedure cannot be used with normally mirrored routine databases if ECP application servers connect to the mirror.
  1. Enact a code freeze and application configuration freeze to disallow application changes during the upgrade, using established procedures at your site, to ensure that routine databases are not modified during the upgrade.
  2. To prevent failover from occurring until the backup is fully upgraded, set no failover.
  3. Perform one of the following two operations:
  4. Clear no failover.
  5. Perform a graceful shutdown of the primary (A). The mirror fails over and the backup (B or C) takes over as primary.
  6. To prevent system A from taking over as primary until fully upgraded, set no failover,
  7. Upgrade system A with the new version of Caché. System A becomes backup.
  8. Upgrade any non-mirrored classes and routines on system A.
  9. If system C became primary and system B was demoted to DR async, upgrade system B and upgrade non-mirrored classes and routines on system B, if any.
  10. For any mirrored routine databases that you removed from the mirror on backup (B or C) before it became the new primary, do the following:
    1. Remove the routine databases from the mirror on system A.
    2. Add the databases to the mirror on the new primary (B or C), then back them up and restore them on the backup (A) and DR async (B, if C is the primary), using the procedures for adding an existing database to a mirror provided in Adding Databases to a Mirror in the “Mirror” chapter of the Caché High Availability Guide. Retain these backups, as they represent the first backups of newly-added mirrored databases; older backups made prior to the upgrade cannot be used to restore these databases in the event of disaster.
  11. Clear no failover.
Major Upgrade with Planned Downtime
If you are upgrading to a new major Caché version and/or performing an application upgrade and have a significant planned maintenance window that obviates the need to minimize mirror downtime, you may prefer to use the following simpler procedure, regardless of whether changes to mirrored databases on the primary are required:
  1. Disallow all user access to the mirror using established practices at your site. Additionally, you must disable any application jobs that would normally start on instance startup. If using Ensemble, disable production auto-start.
  2. Perform a graceful shutdown of the backup (B).
  3. Perform a graceful shutdown of the primary (A).
  4. Upgrade system A with the new version of Caché. System A becomes primary.
  5. Upgrade mirrored classes and routines on system A. If an application upgrade requires further changes to mirrored databases, make these changes.
  6. Upgrade non-mirrored classes and routines on system A, if any.
  7. Perform one of the following two operations:
  8. If system C became primary and system B was demoted to DR async, upgrade system B and upgrade non-mirrored classes and routines on system B, if any.
  9. Restore user access to the mirror.
Upgrading ECP Configurations
In general, ECP application servers should be upgraded before the data servers they connect to. Application servers must have either local routines databases to recompile following upgrade or access to “precompiled” routines database (previously recompiled on a separate system running the target Caché version) on the data server.
Downtime can be minimized using rolling upgrades with users connecting to both upgraded and unupgraded application servers as well as the database server if one of the following is true:
If the new application code can work against old data, but can generate new data structures not understood by the old code, follow this procedure:
  1. Upgrade the application servers on a rolling basis, but do not allow users to connect to an application server once it is upgraded. (Application server capacity will be gradually reduced.)
  2. When enough application servers have been upgraded, restore user access to the upgraded application servers and end all user connections to the remaining (not upgraded) application servers and the data server (if need be), ensuring that users have no access to these systems until you enable it.
  3. Upgrade the remaining application servers, restoring user access to each application server after it is upgraded.
  4. Upgrade the data server and restore user access as needed. (Upgrading the data server will cause a pause in application activity, the length of which will depend on the amount of downtime involved in the upgrade.)
If you have questions or concerns about how to upgrade your ECP configuration, please contact InterSystems Worldwide Customer Support (WRC).
Post-installation Upgrade Tasks
The post-installation tasks required depend on whether you have upgraded to a new major version of Caché or to a maintenance release of the installed version.
Major Version Post-Installation Tasks
After upgrading Caché to a new major version—for example, from 2014.1.0 to 2015.1.0—you must perform the following tasks (if you have not already done so as part of one of the previous procedures in this chapter):
The following tasks may be necessary depending on your environment and the components you use:
Maintenance Release Post-Installation Tasks
After upgrading to a maintenance release of the installed version of Caché—for example, from 2015.1.0 to 2015.1.1—you are not required to recompile any classes or routines. However, if you choose to recompile, it is important to follow the guidance provided in Upgrading From Prior Released Versions in the “General Upgrade Information” chapter of the Caché Upgrade Checklist.
Typically, no changes to external files or clients are required following a maintenance release upgrade.