Women and Technology: sjroscal

I couldn't forget the memory in Shanghai and that's because SUN has been the object of admiration ever since I studied computer in university. How can I explain despair of the day that a shocking news announced. I'm not sure what is the best choice for company and IT market, maybe Sun's technology will be the most advantageous one for our society again. ^^"

DR system is usually configured by similar spec system with source system, it cost lot of money because DR destination system is difficult to be used as mission critical and even as development machine that some system parameters could be changed by various engineers at the same time.
Solaris containers would be the best choice in these cases. Destination system has a container for DR and several containers for other purpose like development. And same application with source will be installed on destination system and DR container in advance like volume manager or oracle, fortunately, zone can access to the vxvm volume easily with vxvm5.0mp3. (volume control is the one of important factor to do DR!) For disk replication, I used shadow image in the test.

To use tape device like robot and drives on logical domain, IO domain with split-pcie bus is required. There isn't virtual pci device until LDom manager 1.2 yet, and don't know it will be configured or not in next, next release.
I would like to check that tape device could run on IO domain of T5240 successfully. T5240 have 2 pcie bus, it means 1 control domain and 1 io domain can be configured at the same time, of course, a san disk as boot disk of io domain and additional network card for control domain are necessary. Tape device runs well on io domain as we expected.

See the short configuration for ldom,

Now, let's check how the number of split-pcie bus changes according to the number of cpus. The number of PCIe buses is equal to the number of CPUSs by system design and it's because the pcie root complex is the part of the cpu silicon.

As a result, max 4 split-pcie bus could be enabled with 4 cpus that is full T5440 configuration, however, what will be happened if only 2cpus are there? We can use only 2 split-pcie bus, of course, all pcie slot are still available after io reconfiguration to reattach each failed pcie slot and onboard device to its nearest available cpu module. (set ioreconfigure=nextboot)

CAM(common array manager) sometimes could be used by CLI and it's faster than BUI.

<sub>#---------------------------------------------
#       Login
#---------------------------------------------
#sscs login -h 10.10.2.154 -u root

#---------------------------------------------
#       register array
#---------------------------------------------
#sscs list storage-system
#sscs register -d  storage-system       #registered after cmd execution
#sscs register -i 10.10.8.124  storage-system
#sscs remove -a STK6140-3 registeredarray

#---------------------------------------------
#       profile & pool
#---------------------------------------------
#sscs create -a STK6140-3 -r 1 -s 512K -h on -n variable -k ANY -H no profile R10
#sscs create -a STK6140-3 -p R10 pool ssc

#---------------------------------------------
#       volume creation: (virtual disk first)
#---------------------------------------------
#sscs create -a STK6140-3 -p ssc -s 55GB -n 16 volume v1
#sscs create -a STK6140-3 -p ssc -s 55GB -v 1 volume v2

#---------------------------------------------
#       Host Group/Host/Initiator
#       (fcinfo hba-port)
#---------------------------------------------
#sscs create -a STK6140-3 hostgroup ssc
#sscs create -a STK6140-3 -g ssc host jana
#sscs create -a STK6140-3 -w 210000e08b091e61 -h jana -o solaris initiator jana-1
#sscs create -a STK6140-3 -w 210000e08b291e61 -h jana -o solaris initiator jana-2

#---------------------------------------------
#       map/unmap
#---------------------------------------------
#sscs map -a STK6140-3 -P readwrite -h jana volume a1
#sscs unmap -a STK6140-3 volume a01

#---------------------------------------------
#       delete
#---------------------------------------------
#sscs delete -a STK6140-3 volume a01

#---------------------------------------------
#       List
#---------------------------------------------
#sscs list -a STK6140-3 controller</sub>

We couldn't run shell script directly on XSCF to schedule dynamic reconfiguration automatically on fixed time or according to the cpu usage. That's because XSCF is the management S/W, not OS.
Automatic resource management is the one of the most important requirements for datacenter, and it can be configured by ssh remote login and the expect for not requiring password prompt. Also SunMC would be used together when we need more visual configuration.

================
vi run_dr.exp
================
#!/usr/local/bin/expect
spawn ssh -l platadm 10.10.3.10 "deleteboard -y -c unassign 07-0;addboard -y -c assign -d 3 07-0"
#spawn ssh -l platadm 10.10.3.10 "addboard -y -c configure -d 2 07-0"
set password root
expect "[email protected]'s password:"
send "$password\r"
expect eof

================
vi run_dr.sh
================
/usr/local/bin/expect -f /scripts/dr/run_dr.exp