Hot-plug mirrored memory, Hot-plug mirrored, Memory" on – HP ProLiant DL580 G3 Server User Manual
Page 100
100
HP ProLiant DL580 Generation 3 Server User Guide
•
The minimum valid online spare configuration for a memory board requires
at least one bank of dual-rank DIMMs or two banks of single-rank DIMMs.
If the server does not meet these requirements, an error message appears
during POST and the server defaults to Advanced ECC ("Advanced ECC
memory" on page
•
The server automatically configures the optimal online spare solution.
•
Hot-plug operations are not supported.
HP recommends the following configurations. These configurations result in
optimal use of memory. Other configurations are valid, but do not result in the
maximum amount of installed memory being available to the operating system.
•
If only single-rank DIMMs are used on a memory board, all DIMMs should
be of the same size on that memory board.
•
If only dual-rank DIMMs are used on a memory board, all DIMMs should be
of the same size on that memory board.
•
If a mixture of single- and dual-rank DIMMs are used on a memory board,
the dual-rank DIMMs should be twice the size of any single-rank DIMM.
After installing DIMMs, use RBSU to configure the server ("Configuring the
memory" on page
) for online spare memory support.
Hot-plug mirrored memory
Hot-plug mirrored memory provides a higher level of memory protection than
either Advanced ECC ("Advanced ECC memory" on page
) or online spare
memory (on page
). With hot-plug mirrored memory, the server is protected
against uncorrectable memory errors that would otherwise result in server failure.
Hot-plug mirrored memory allows the server to keep two copies of all memory
data on separate memory boards.
If an uncorrectable error is encountered, the proper data is retrieved from the
memory board that did not fail. In addition, hot-plug mirrored memory allows
failed or degraded DIMMs to be replaced while the server is running without
requiring server downtime. The memory board with the failed DIMMs can be
removed, failed DIMMs replaced, and the board re-inserted into the server
without any interruption to the operating system.