How to perform a minimal downtime upgrade from 5.0 / 5.0MP3 to 6.0 or 6.0.1 of Storage Foundation HA (VCS, VxFS, and VxVM)

book

Article ID: 100027648

calendar_today

Updated On:

Resolution

1. Freeze the service groups.
a) Make the cluster configuration read / write.
# haconf -makerw
b) Persistent freeze each and all cluster systems (perform this step for each cluster node).
# hasys -freeze –persistent
c) Save and dump the cluster configuration.
# haconf -dump -makero

2. Upgrade the VCS portion of the product on all systems.  
a) Stop VCS while leaving the service groups online, running outside VCS.
# hastop -all -force
b) Verify on each node that port h (HAD - high availability daemon) is no longer running.
This is an example of an expected output:
# gabconfig -a
GAB Port Memberships
===============================================================
Port a gen   22590b membership 01

c Update VCS only on all nodes of the cluster first to 5.0MP3 ( the update must be done for all cluster nodes at the same time, or the cluster nodes will not join due to version mismatch errors).
# cd /cluster_server
# ./installvcs  
(enter all cluster node names when prompted)

) Update VCS only on all nodes of the cluster first to 6.0 the update must be done for all cluster nodes at the same time, or the cluster nodes will not join due to version mismatch errors).
# cd /cluster_server
# ./installvcs  -rollingupgrade_phase2
(enter all cluster node names when prompted)

Note that the installation script may indicate a reboot upon completion, but no reboot for any of the clustered systems is required at this point in time.

3. Start VCS on all nodes to verify functionality and switch service groups to one node.
a) Start VCS on each node.
# hastart
b) Verify that the cluster nodes joined successfully by running the command on each node (port a, and h should have a single line in the output).
This is an example of an expected output:
# gabconfig -a      
GAB Port Memberships
===============================================================
Port a gen   22590b membership 01
Port h gen   225916 membership 01

c) Verify the status of the service groups in the cluster.
# hastatus –sum
---At this point VCS on both nodes should be at version 6.0
Check with pkginfo –l VRTSvcs

4. Switch all the service groups to one node (call this nodeA) and persistent freeze the systems.
a) Make the cluster configuration read / write.
# haconf -makerw
b) Persistent unfreeze each and all cluster systems (perform this step for each cluster node).
# hasys -unfreeze –persistent
c) Using the output from step 3c, switch all service groups from nodeB to nodeA.
# hagrp -switch -to
d) Persistent freeze each and all cluster systems (perform this step for each cluster node).
# hasys -freeze -persistent
e) Save and dump the cluster configuration.
# haconf -dump -makero

5. Update VxVM and VxFS on nodeB and reboot.
At this time, VCS is updated on all nodes.
The product installation scripts will only update the lower versions of the installed products.
There is a known issue with the 5.0MP3 installation script.
Executing any of the installation scripts more than one time (as is required for a minimal downtime update) may trigger the issue. There is a hotfix to resolve this issue.
The link for the hotfix is in the Related Documents section near the bottom of this document.
The perl script in this hotfix may be utilized with any of the installation scripts included with the product (for example: installer, installmp, installsf, installvcs, and installvm).
Please download the CPI hotfix, copy the perl script in the hotfix to /var/tmp and utilize the perl script from the hotfix as required by the command below.
a) Update VxFS to 5.0MP3 on nodeB.
# ./installmp -require /var/tmp/rem_drv_patch.pl
b) Upon completion of the script, reboot is indicated.
# reboot
a) Update VxFS and VxVM  to 6.0 on nodeB.
# ./ installer -rollingupgrade_phase1
b) Upon completion of the script, reboot is indicated.
# reboot

NodeB upgrade is now complete.
Check modinfo | grep vx

6. Unfreeze the systems and then switch the service groups from nodeA to nodeB. Then persistent freeze the systems.
a) Make the cluster configuration read / write.
# haconf -makerw
b) Unfreeze the systems.
# hasys -unfreeze -persistent  
c) Switch all groups to nodeB.
# hagrp -switch -to
d) Freeze persistent all nodes.
# hasys -freeze -persistent
e) Save and dump the cluster configuration.
# haconf -dump -makero

7. Update VxVM and VxFS on nodeA and reboot.
a) Update VxFS to 5.0MP3 on nodeB.
# ./installmp -require /var/tmp/rem_drv_patch.pl
b) Upon completion of the script, reboot is indicated.
# reboot
a) Update VxFS and VxVM  to 6.0 on nodeB.
# ./ installer -rollingupgrade_phase1
b) Upon completion of the script, reboot is indicated.
# reboot

NodeA upgrade is now complete.
Check modinfo | grep vx

8.  Unfreeze and move groups (if desired) to their original locations.
 a) Make the cluster configuration read / write.
# haconf -makerw
b) Unfreeze the systems.
# hasys -unfreeze -persistent
c) Switch any desired groups.
# hagrp -switch -to
d) Save and dump the cluster configuration.
# haconf -dump -makero
e) Verify all service groups are online.
# hastatus -sum

The upgrade is complete. The Cluster is now at 6.0.


Issue/Introduction

The purpose of this document is to provide an example of the steps required to update a 5.0/ 5.0MP3 on a 2 node cluster to 6.0  or 6.0.1 with minimal downtime. The process given is an example only.
Different configurations / environment may require changes to the steps that are detailed below.
Additionally, this example does not take into account boot disk encapsulation.
 
Note: There is a known issue with multiple upgraded for 5.0MP3 – this requires the CPI Hot fix.
Instructions for usage:
Solution:
Veritas is providing a new Perl module as a workaround to this installation problem.
 
Workaround Steps:
      1. Copy the Perl module rem_drv_patch.pl to /tmp
      2. Usage:
          ./installmp -require /tmp/rem_drv_patch.pl

During the update process, the Volume Manager and VxFS Filesystem components will remain at a lower level than VCS as the process is completed.
Storage Foundation with Cluster File System (CFS) and / or Oracle RAC require all Veritas product components to be at the same level.   Therefore, this process does not apply to CFS or RAC installations.
The process may be modified for use with clusters that have more than 2 nodes.

Powered by Wolken Software