Filter
Top Products
150 Chapter 7
Recovering from failures
Abnormal system shutdowns
To calculate an appropriate size for a V2500/V2600 SCA crash dump
volume, estimate that you will need at most the following amount of
space: the total amount of physical memory in the system, plus space to
allow for dump headers and tables, minus the amount of memory
dedicated as CTI cache, and minus the amount of memory that is kernel
text replicated across cabinets.
The total amount of space required depends on whether a full dump or a
selective dump is performed, and whether the crash dump is compressed
or uncompressed.
Configuration criteria
There are three main criteria to consider when making decisions
regarding how to configure system dumps. The criteria are:
• System recovery time—Get the system back up as soon as possible
• Crash information integrity—Capture the correct information
• Disk space needs—Conserve available disk space
System recovery time
To get the system back up and running as soon as possible, consider the
following factors:
•Dump level
• Compressed save vs. noncompressed save
• Using a device for both paging and dumping
• Partial save
These factors are discussed in the following sections.
Dump level
With HP-UX 11.0 the operator can select three levels of core dumps: no
dump, selective dump, or full dump.
Selective dump causes only the selected memory pages to get dumped,
see the crashconf(1M) man page for more information.
NOTE In some specific cases, HP-UX will override the selective dump and
request a full dump. The operator is given ten seconds to override HP-UX
and continue with a selective dump.