Technical data

Table 5–4: Adding a Memory Channel Interconnect or Upgrading from a
Dual, Redundant MC1 Interconnect to MC2 Interconnects (cont.)
Step
Action Refer to:
• The last member system has had its second MC1
adapter replaced with an MC2 adapter.
• The cluster is operational.
• All MC2 adapters are jumpered for 512 MB (and you
need to utilize 512 MB address space).
On one member system, use the sysconfig command
to reconfigure the Memory Channel kernel subsystem to
initiate the use of 512 MB address space. The configuration
change is propagated to the other cluster member systems:
/sbin/sysconfig -r rm rm_use_512=1
______________________ Note ______________________
After the configuration change is propagated to the other member
systems, you can reboot any member system and the 512 MB address
space is still in effect.
If you use the sysconfig command to promote address space to
512 MB and inadvertently leave an MC2 adapter jumpered for 128
MB, then reboot that system, it will not rejoin the cluster. When the
system with the Memory Channel adapter jumpered for 128 MB is
shut down, and the TruCluster software running on the remaining
cluster member systems discover that all operational Memory Channel
adapters are jumpered for 512 MB, because address space has
been promoted to 512 MB, the active rail will use 512 MB address
space. A system jumpered for 128 MB cannot join the cluster. The
startup error message on the system jumpered for 128 MB follows:
panic: MC2 adapter has too little memory
If you have used the sysconfig command to promote Memory Channel
address space to 512 MB, you may need to know the actual address space
being used by a logical rail. Use the dbx debugger utility as follows to
determine:
• Logical size (in 8-KB pages) of a rail
• Physical size (J3 jumper setting) for physical rails
# dbx -k /vmunix
(dbx) p rm_log_rail_to_ctx[0]->mgmt_page_va->size
1
16384 2
(dbx) p rm_adapters[0]->rmp_prail_va->rmc_size 3
{
[0] 65536
4
[1] 0
5–18 Setting Up the Memory Channel Cluster Interconnect