2-42
Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
Compaq Confidential – Need to Know Required
Writer:
Bryan Hicks
Project:
Compaq ProLiant Clusters HA/F100 and HA/F200 Administrator Guide
Comments:
Part Number:
380362-003
File Name:
c-ch2 Designing the Compaq ProLiant Clusters HAF100 and HAF200.doc
Last Saved On:
8/24/00 12:00 PM
Failover Threshold and Failover Period
The failover threshold and failover period are similar to the restart values. The
failover threshold defines the maximum number of times per failover period
that the Microsoft clustering software attempts to fail over a cluster group. If
the cluster group exceeds the failover threshold in the allotted failover period,
the group is left on its current node, in its current state, whether that is online,
offline, or partially online.
The failover threshold and failover period prevents a cluster group from
bouncing back and forth between servers. If a cluster group is so unstable that
it cannot run properly on either cluster node, it will eventually be left in its
current state on one of the nodes. The failover threshold and period determine
the point at which the decision is made to leave the cluster group in its current
state.
The following example illustrates the relationship between the restart
threshold and period and the failover threshold and period.
Assume you have a cluster group (Group1) that is configured to have a
preferred server (Server1). If Group1 encounters an event that forces it offline,
MSCS attempts to restart the resource. If Group1 cannot be restarted within
the limits of the restart threshold and period, the Microsoft clustering software
attempts to fail over Group1 to Node 2. If the failover threshold for Group1 is
set to 10 and the failover period is set to 3 (hours), the Microsoft clustering
software will fail over Group1 as many as 10 times in a 3-hour period. If a
failure is still forcing Group1 offline after three hours, the Microsoft clustering
software will no longer attempt to fail over the group.
Failover of Directly Connected Devices
Devices that are physically connected to a server cannot move to the other
cluster node. Therefore, any applications or resources dependent on these
devices may be unable to restart on the other cluster node. Examples of
direct-connect devices include printers, mainframe interfaces, modems, fax
interfaces, and customized input devices such as bank card readers.
For example, if a server is providing print services to users, and the printer is
directly connected to the parallel port of the server, there is no way to switch
the physical connection to the other server, even though the print queue and
spooler can be configured to fail over. The printer should be configured as a
true network printer and connected to a hub that is accessible from either
cluster node. In the event of a server failure, not only will the print queue and
spooler fail over to the other server, but physical access to the printer will be
maintained.
Содержание ProLiant Clusters HA/F100
Страница 8: ......