HP AE370A - Brocade 4Gb SAN Switch 4/12, Administrator'S Manual

Page: 402 / 447

386 Configuring the PID format
It is also important to understand how multipathing software reacts when one of the two fabrics is taken
offline. If the time-outs are set correctly, the failover between fabrics should be transparent to the users.
You should use the multipathing software to manually fail a path before starting maintenance on that
fabric.
4. Perform empirical testing.
Empirical testing might be required for some devices, to determine whether they bind by PID. If you are
not sure about a device, work with the support provider to create a test environment.
Create as close a match as practical between the test environment and the production environment,
and perform an update using the procedure in ” Online update ” on page 386.
Devices that bind by PID are unable to adapt to the new format, and one of three approaches must be
taken with them:
• A plan can be created for working around the device driver’s limitations in such a way as to allow
an online update. See the Detailed Procedures section for examples of how this could be done.
• The device can be upgraded to drivers that do not bind by PID.
• Downtime can be scheduled to reset the device during the core PID update process, which generally
allows the mapping to be rebuilt.
If either of the first two options are used, the procedures should again be validated in the test
environment.
Determine the behavior of multipathing software, including but not limited to:
• HBA time-out values
• Multipathing software time-out values
• Kernel time-out values
Planning the update procedure
Whether it is best to perform an offline or online update depends on the uptime requirements of the site.
• An offline update that all devices attached to the fabric be offline.
• With careful planning, it should be safe to update the core PID format parameter in a live, production
environment. This requires dual fabrics with multipathing software. Avoid running backups during the
update process, as tape drives tend to be very sensitive to I/O interruption. The online update process
is only intended for use only in uptime-critical dual-fabric environments, with multipathing software
(high-uptime environments should always use a redundant fabric SAN architecture). Schedule a time for
the update when the least critical traffic is running.
All switches running any version of Fabric OS 3.1.2 and later or 4.2.0 and later are shipped with the Core
Switch PID Format enabled, so it is not necessary to perform the PID format change on these switches.
Migrating from manual PID binding (such as persistent binding on an HBA) to manual WWN binding and
upgrading drivers to versions that do not bind by PID can often be done before setting the core PID format.
This reduces the number of variables in the update process.
Online update
The following steps are intended to provide SAN administrators a starting point for creating site-specific
procedures.
1. Back up all data and verify backups.
2. Verify that the multipathing software can automatically switchover between fabrics seamlessly. If there is
doubt, use the software’s administrative tools to manually disassociate or mark offline all storage
devices on the first fabric to be updated.
3. Verify that I/O continues over the other fabric.
4. Disable all switches in the fabric to be updated, one switch at a time, and verify that I/O continues over
the other fabric after each switch disable.
5. Change the PID format on each switch in the fabric.
6. re-enable the switches in the updated fabric one at a time. In a core/edge network, enable the core
switches first.