Before upgrading your clustered filers, verify that the cluster is properly configured by using the Cluster Configuration Checker. Identify and resolve any cluster configuration issues before continuing with the upgrade.
The Cluster Configuration Checker is a Perl script that detects errors in the configuration of a pair of clustered NetApp filers. The script is available on the NOW site at http://now.netapp.com/NOW/download/tools/cf_config_check/.
See also the chapter on cluster management in the Storage Management Guide.
The following methods are available for upgrading cluster partners:
Depending on the nature of the upgrade, nondisruptive cluster upgrades ensure better filer availability than single system or standard cluster upgrades.
When planning for your upgrade, evaluate the advantages and disadvantages of each method. Each method is described in this section.
In a nondisruptive upgrade (also called a rolling upgrade), clustering technology is used to enable a takeover filer to assume the functions of the "failed" partner while it is being upgraded. There is a takeover and giveback operation for each cluster node. Because the cluster partner node fulfills service requests during the "failed" filer's upgrade, no disruption in service is experienced by the clients.
In addition, because the takeover filer assures continuous availability of the "failed" filer's disks, more extensive upgrades requiring a system halt--such as system firmware upgrades and hardware adapter replacements--can be performed without disrupting services based on stateless protocols.
Network Appliance recommends nondisruptive clustered upgrades to minimize service disruption when one or more of the following is being performed:
Before undertaking a nondisruptive upgrade procedure, Network Appliance recommends that you do the following:
Nondisruptive upgrades require more planning and procedural steps than standard upgrades. Make sure to thoroughly review this document before starting a nondisruptive upgrade. See especially Appendix D, Optimizing Service Availability During Upgrades.
You can verify that your clustered filers are properly configured by running the Cluster Configuration Checker as described in Before upgrading clustered filers.
Configure client timeout settings to optimize availability as described in Service and protocol considerations.
You should not use nondisruptive upgrades in the following circumstances:
Because CIFS is session-oriented, sessions must be terminated before upgrade procedures to prevent data loss.
Because these protocols are session-oriented, outstanding sessions must complete, and these services must be disabled in order to use nondisruptive upgrades.
Disk firmware upgrades automatically take disks offline until the upgrade is complete. Services and data are unavailable until they are back online.
On diskette-based clustered filers, the takeover and giveback operations are much slower than on CompactFlash-based filers and possibly slower than rebooting.
For these conditions, Network Appliance recommends standard upgrades.
Note
Nondisruptive upgrades cannot be employed for upgrades between release families--standard upgrades must be used.
In a standard upgrade, the cluster is disabled and each node is updated. When the cluster is disabled, each node behaves as a single-node (non-clustered) filer; in other words, system services associated with that node are interrupted for as long as it takes the system to reboot.
In general, standard cluster upgrades are most appropriate if you schedule regular downtime for your filers. Other maintenance upgrades, such as system firmware and hardware, can also take place when the cluster is disabled.
Standard cluster upgrades must be employed for upgrades between Data ONTAP release families.
Copyright © 1994-2004, Network Appliance, Inc. All rights
reserved.
Part No. 215-01027
File: upgrade4.htm
Last updated: Mon Jul 12 17:06:05 2004