Database Reference
In-Depth Information
•
Also, it's important to refer to the OS watcher logs to identify resource consumption on the
node during or just before node eviction
•
OS Watcher log files—All OS Watcher log files are stored under the archive log directory under
OS Watcher directory. It contains significant information about resource utilization, such as
CPU, memory, top output, swap (page in/out), etc., information. The statistical information
recorded in the log files help in identifying if the system was overloaded or starving for
resources (CPU, memory)
Rebootless Stop/Fencing
With the incorporation of
rebootless fencing
behavior with 11gR2, a node actually under certain circumstances
won't reboot completely; rather, it will attempt to stop the GI on the node gracefully to avoid complete node eviction.
For example, when the GI home or
/var
file system runs out of space on the device,
cssd/crsd,evmd
processes
become unhealthy, in contrast to pre-11gR2 behavior, and this prevents a total node reboot scenario. Restarting
the individual stack resources after resolving the underlying issues in the context is recommended to bring the
cluster stack back to a healthy situation. However, if the
rebootless fencing
fails to stop one or more resources, the
rebootless fencing
will fail and the
cssdagent
or
cssdmonitor
will favor and trigger complete node eviction and
will perform a fast reboot. In this context, a warning message of resource failure will be written to the
alert.log
file.
Extended Distance (Stretch) Clusters—Synopsis, Overview,
and Best Practices
Extended distance (stretch) clusters (Figure
7-10
) are a very rare species and justifiably so. RAC clusters are generally
architected to reside within a single geographical location. Stretch clusters are generally defined as RAC clusters with
nodes spanning over multiple data centers, resulting in a geographically spread-out private cluster interconnect.
Figure 7-10.
Example configuration: Extended distance (stretch) RAC cluster
