Top
PRIMECLUSTER Global Link Services Configuration and AdministrationGuide 4.3Redundant Line Control Function

4.5.3 Replacement/Expansion PHP (PCI Hot Plug)

This section explains a procedure of replacing or adding a PCI card for GLS in a PCI Hot Plug (PHP) environment.

Compatibility of PHP with each mode is shown in the table below.

Table 4.2 Compatibility of PHP with each mode

PHP(PCI Hot Plug) operation

Fast switching mode

NIC switching mode

GS/SURE linkage mode

Replacement (Redundant system)

A

A

A

Extension (Non-redundant system)

A

A

A

Extension (Redundant system)

A

X

A

[PHP Support] A: Supported X: Not supported

Note

Replacement and expansion of PHP (PCI Hot Plug) is allowed only when the system is running in a multiple user mode.

4.5.3.1 Replacement of PCI card on redundant system

In Fast Switching and NIC Switching mode, it is possible to replace the redundant NIC without stopping network communication.

Note

For NIC Switching mode, it is required to stop the transfer path monitoring function and standby patrol function.
For GS/SURE linkage mode, it is required to deactivate the virtual interface.

The following is a procedure of replacing redundant system.

1) Specify the replacing PCI card

An interface on the PCI card to be replaced can be identified from the warning messages output to the console (eg. igb1).

Note

In Solaris 11 or later, Solaris 10 9/10 or later releases, or in the environment where 142909-17 or later patch is applied in Solaris 10

When using PHP, execute the following command to enable the hotplug service. For details, refer to "Service Manual" according to your server.

# svcadm enable hotplug

2) Disconnect from redundant system

In order to remove the PCI card from the redundant system for replacement, please execute the following command with the interface name obtained in the procedure "1) Specify the replacing PCI card". On the redundant system of standby mode, when an online communication path is disconnected, a standby communication path will be online communication path automatically.

Fast switching mode
NIC switching mode

The following procedure is required depending on the status of a NIC (PCI card) built into the target system board for replacement.

  • If OS is Solaris 11 and a NIC is the primary interface:
    Settings of the IP address must be deleted before replacing the system board. If a NIC is the operation NIC, the NIC switching is required manually before replacing the system board.

  1. Checking the primary interface name

    Check the primary interface name. The primary interface name is displayed at the beginning of "Interface List" field. In this example, it is "net1."

    # /opt/FJSVhanet/usr/sbin/hanetconfig print
    [IPv4,Patrol]
    
     Name        Hostname        Mode MAC Adder/Phys ip Interface List
    +-----------+---------------+----+-----------------+---------------------------+
     sha0        192.168.70.1     d   192.168.70.2      net1,net2
     sha1        -                p   00:00:00:00:00:00 sha0
    
    [IPv6]
    
     Name        Hostname/prefix                   Mode Interface List
    +-----------+---------------------------------+----+---------------------------+
    

    Check the operation NIC name. In this example, it is "net1."

    # /opt/FJSVhanet/usr/sbin/dsphanet
    [IPv4,Patrol]
     Name       Status   Mode CL  Device
    +----------+--------+----+----+------------------------------------------------+
     sha0       Active    d   ON   net1(ON),net2(OFF) 
     sha1       Active    p   OFF  sha0(ON) 
    [IPv6]
     Name       Status   Mode CL  Device
    +----------+--------+----+----+------------------------------------------------+
  2. Stopping the HUB monitoring function and the patrol monitoring function

    # /opt/FJSVhanet/usr/sbin/hanetpoll off <Return>
    # /opt/FJSVhanet/usr/sbin/stpptl -n sha1 <Return>
  3. Switching the physical interfaces

    To execute this command, a NIC for replacement must be the operation NIC.

    # /opt/FJSVhanet/usr/sbin/hanetnic change -n sha0 <Return>
  4. Deleting IP address setting

    To execute this command, OS must be Solaris 11 and NIC for replacement must be the primary interface.

    # /usr/sbin/ifconfig net2 <Return>
    # /usr/sbin/ipadm delete-ip net1 <Return>

    Note

    To return to the multiplex configuration, the IP address and the netmask of the interface specified by the ipadm delete-ip command must be set.

    Note the output result of the ifconfig command.

GS/SURE linkage mode
# /opt/FJSVhanet/usr/sbin/hanetnic delete -n sha0 -i sha2 <Return>

3) Disconnect the PCI card

Specify the interface name (eg, igb1) identified in the procedure "1) Specify the replacing PCI card." to "inst2comp" command to obtain the PCI bus slot "Ap_Id".

Specify the "Ap_Id" obtained above as an argument to "cfgadm"(1M) command, and confirm that the slot status of the PCI card to be disconnected is "connected configured".

# cfgadm BB#0-PCI#0
Ap_Id                          Type        Receptacle    Occupant      Condition
BB#0-PCI#0                     etherne/hp  connected     configured    ok

Execute the uncfgadm(1M) command with specifying unconfigure to cancel the configuration of the PCI card at Ap_Id. After that, execute the cfgadm(1M) command to check if the slot status is changed to "connected unconfigured".

# cfgadm -c unconfigure BB#0-PCI#0 <Return>
# cfgadm BB#0-PCI#0 <Return>
Ap_Id                          Type         Receptacle   Occupant     Condition
BB#0-PCI#0                     unknown      connected    unconfigured unknown

Execute the uncfgadm(1M) command with specifying disconnect to disconnect the PCI card from Ap_Id.

After that, execute the cfgadm(1M) command to check if the slot status is changed to "disconnected unconfigured".

# cfgadm -c disconnect BB#0-PCI#0 <Return>
# cfgadm BB#0-PCI#0 <Return>
Ap_Id                          Type        Receptacle    Occupant      Condition
BB#0-PCI#0                     unknown     disconnected  unconfigured  unknown

To indicate the slot position for replacement, specify the obtained "Ap_Id" to "cfgadm" command and blink the ATTENTION LED.

# cfgadm -x led=attn,mode=blink BB#0-PCI#0 <Return>

4) Replace the PCI card

The PCI card disconnected in the procedure "3) Disconnect the PCI card" is replaced with a new one. Our customer support staff does this for you.

5) Connect the PCI card

Execute the cfgadm(1M) command with specifying connect to connect the new PCI card to Ap_Id.

After that, execute the cfgadm(1M) command to check if the slot status is changed to "connected unconfigured".

# cfgadm -c connect BB#0-PCI#0 <Return>
# cfgadm BB#0-PCI#0 <Return>
Ap_Id                          Type         Receptacle   Occupant     Condition
BB#0-PCI#0                     unknown      connected    unconfigured unknown

Execute the cfgadm(1M) command with specifying connect to connect the new PCI card to Ap_Id.

After that, execute the cfgadm(1M) command to check if the slot status is changed to "connected unconfigured".

# cfgadm -c configure BB#0-PCI#0 <Return>
# cfgadm BB#0-PCI#0 <Return>
Ap_Id                          Type        Receptacle    Occupant      Condition
BB#0-PCI#0                     etherne/hp  connected     configured    ok

6) Connect to redundant system

In order to connect the new PCI card to a redundant system, please execute the following commands with the interface name identified in the procedure "1) Specify the replacing PCI card".

The interface names described in the following procedure vary depending on the environment.

For Solaris 11, the default interface name is netX (X means the instance number).

Fast switching mode

When setting configuration for fast-switching mode, execute the following commands to activate the added interfaces in advance:

NIC switching mode

The following procedure is required depending on the status of a NIC (PCI card) built into the target system board for replacement.

  • If OS is Solaris 11 and NIC is the primary interface:
    Settings of the IP address must be deleted before replacing the system board. After replacing the system board, switch the NIC manually, and reset the IP address.

  1. Switching the physical interface and resetting the IP address

    To execute this command, OS must be Solaris 11 and a NIC built into the replaced system board must be the primary interface. Otherwise, move to Step 2.

    • If the IPv4 address is used

      # /usr/bin/touch /var/opt/FJSVhanet/tmp/disable_watchif <Return>
      # /opt/FJSVhanet/usr/sbin/hanetnic change -n sha0 <Return>
      # /usr/sbin/ipadm delete-ip net1 <Return>
      # /usr/sbin/ipadm create-ip net1 <Return>
      # /usr/sbin/ipadm create-addr -T static -a ipaddress/netmask net1/v4 <Return>
      # /usr/bin/rm -f /var/opt/FJSVhanet/tmp/disable_watchif <Return>
    • If the IPv6 address is used

      # /usr/bin/touch /var/opt/FJSVhanet/tmp/disable_watchif <Return>
      # /opt/FJSVhanet/usr/sbin/hanetnic change -n sha0 <Return>
      # /usr/sbin/ipadm delete-ip net1 <Return>
      # /usr/sbin/ipadm create-ip net1 <Return>
      # /usr/sbin/ipadm create-addr -T addrconf net1/v6 <Return>
      # /usr/bin/rm -f /var/opt/FJSVhanet/tmp/disable_watchif <Return>
  2. Starting the HUB monitoring function and the patrol monitoring function

    # /opt/FJSVhanet/usr/sbin/strptl -n sha1 <Return>
    # /opt/FJSVhanet/usr/sbin/hanetpoll on <Return>
  3. Switching the physical interfaces

    Switch the physical interfaces as necessary.

    # /opt/FJSVhanet/usr/sbin/hanetnic change -n sha0 <Return>
GS/SURE linkage mode
# /opt/FJSVhanet/usr/sbin/hanetnic add -n sha0 -i sha2 <Return>

7) Switch back the redundant path

Please switch back an online communication path if needed.

4.5.3.2 Extension of PCI cards with new redundant system

By adding a new PCI card to a non-redundant system, it is possible to create a redundant system.

The following is the procedure of adding a new PCI card to a non-redundant system.

Note

In Solaris 11 or later, Solaris 10 9/10 or later releases, or in the environment where 142909-17 or later patch is applied in Solaris 10

When using PHP, execute the following command to enable the hotplug service. For details, refer to "Service Manual" according to your server.

# svcadm enable hotplug

1) Add PCI cards

a.
Before adding a PCI card, please save the output of "prtpicl"(1M) command.

# prtpicl -v > /tmp/prtpicl.pre <Return>

b.
Check the status of the slot.

An example of adding a PCI card to BB#1-PCI#0 is shown in this section.

Please confirm that the status of the PCI slot where a PCI card is added is "empty unconfigured" by using "cfgadm" (1M) command.

# cfgadm BB#1-PCI#0 <Return>
Ap_Id                          Type        Receptacle    Occupant      Condition
BB#1-PCI#0                     unknown     empty         unconfigured  unknown

c.
To indicate the slot position for expansion, specify the "Ap_Id" identified in the procedure 1)b to "cfgadm" command and blink the ATTENTION LED.

# cfgadm -x led=attn,mode=blink BB#1-PCI#0 <Return>

d.
Add a PCI card after the POWER LED of the target PCI bus slot has turned off.
This operation is performed by our customer support.

e.
Please confirm that the PCI slot status which extended PCI cards is "disconnected unconfigured" by using cfgadm (1M) command.

# cfgadm BB#1-PCI#0 <Return>
Ap_Id                          Type        Receptacle    Occupant      Condition
BB#1-PCI#0                     unknown     disconnected  unconfigured  unknown

2) Connect PCI cards

Execute the cfgadm(1M) command with specifying connect to connect the new PCI card to Ap_Id.
After that, execute the cfgadm(1M) command to check if the slot status is changed to "connected unconfigured".

# cfgadm -c connect BB#1-PCI#0 <Return>
# cfgadm BB#1-PCI#0 <Return>
Ap_Id                          Type         Receptacle   Occupant     Condition
BB#1-PCI#0                     unknown      connected    unconfigured unknown

Execute the cfgadm(1M) command with specifying connect to connect the added PCI card to Ap_Id.
After that, execute the cfgadm(1M) command to check if the slot status is changed to "connected unconfigured".

# cfgadm -c configure BB#1-PCI#0 <Return>
# cfgadm BB#1-PCI#0 <Return>
Ap_Id                          Type        Receptacle    Occupant      Condition
BB#1-PCI#0                     etherne/hp  connected     configured    ok

3) Connect equipment to PCI cards

The extended PCI card is connected with network equipment by the cable.

4) Setup driver

The driver configuration is added by following operations.

a.
The interface name is investigated in order to configure drivers and high layer products.
Please save the result of the prtcicl (1M) command, and obtain the difference information between current result and the result taken at "1) a." Then, the driver instance number for the extended PCI card is obtained.
In the following example, since the instance number is 1, it can be determined that the interface name of the extended PCI card is "igb1."

# prtpicl -v > /tmp/prtpicl.post <Return>
# diff /tmp/prtpicl.pre /tmp/prtpicl.post | more <Return>
            :
> :devfs-path    /pci@8900/pci@4/pci@0/pci@0/network@0
> :driver-name   igb
> :binding-name  pciex8086,1521
> :bus-addr      0
> :instance      1
> :_class        network
> :name          network
            :

b.
Confirm that the interface name that is obtained from the above operation matches the one that has been added to the PCI bus slot.

c.
The configuration of each driver is added.
Please refer to each driver manual for details.

5) Setup redundant system

Activate the virtual interface after configuring Fast Switching, NIC Switching, or GS/SURE linkage mode. System reboot is not required after configuring each mode.

When configuring Fast Switching mode, the added interface "igb1(net1)" must be activated preliminary by the following command.

The interface names described in the following procedure vary depending on the environment.

For Solaris 11, the default interface name is netX (X means the instance number).

Note

[For Solaris 10 environment]

For Fast switching mode (IPv4), the IP address specified in this section must also be defined in /etc/inet/hosts and /etc/hostname.igb1.
For Fast switching mode (IPv6), create /etc/hostname6.igb1 as an empty file.
Unless these are configured, when the system reboots, the virtual interface for Fast Switching mode cannot be activated.

4.5.3.3 Extension of PCI cards to redundant system

It is possible to extend a PCI card to the redundant system.

Note

  • In NIC Switching mode, it is not possible to add a new interface to a redundant system.
    Also, when adding a new interface in GS/SURE linkage mode, it is first required to deactivate the virtual interface for GS/SURE linkage mode and then add a new interface.

  • When using PHP in Solaris 11 or later, Solaris 10 9/10 or later releases, or in the environment where 142909-17 or later patch is applied in Solaris 10, execute the following command to enable the hotplug service. For details, refer to "Service Manual" according to your server.

    # svcadm enable hotplug

The following is the procedure of extending PCI card to the redundant system.

1) Add PCI cards

a.
Before extending the PCI card, please save the result (current configuration information) of the prtpicl (1M) command.

# prtpicl -v > /tmp/prtpicl.pre <Return>

b.
Check the status of the slot.
An example of adding a PCI card to BB#1-PCI#0 is shown in this section. Please confirm that the status of the PCI slot where a PCI card is added is "empty unconfigured" by using "cfgadm" (1M) command.

# cfgadm BB#1-PCI#0 <Return>
Ap_Id                          Type        Receptacle    Occupant      Condition
BB#1-PCI#0                     unknown     empty         unconfigured  unknown

c.
To indicate the slot position for expansion, specify the "Ap_Id" identified in the procedure 1) b to "cfgadm" command and blink the ATTENTION LED.

# cfgadm -x led=attn,mode=blink BB#1-PCI#0 <Return>

d.
Add a PCI card after the POWER LED of the target PCI bus slot has turned off. This operation is performed by our customer support.

e.
Please confirm that the PCI slot status which extended PCI cards is "disconnected unconfigured" by using cfgadm (1M) command.

# cfgadm BB#1-PCI#0 <Return>
Ap_Id                          Type        Receptacle    Occupant      Condition
BB#1-PCI#0                     unknown     disconnected  unconfigured  unknown

2) Connect PCI cards

Execute the cfgadm(1M) command with specifying connect to connect the new PCI card to Ap_Id.

After that, execute the cfgadm(1M) command to check if the slot status is changed to "connected unconfigured".

# cfgadm -c connect BB#1-PCI#0 <Return>
# cfgadm BB#1-PCI#0 <Return>
Ap_Id                          Type         Receptacle   Occupant     Condition
BB#1-PCI#0                     unknown      connected    unconfigured unknown

Execute the cfgadm(1M) command with specifying connect to connect the added PCI card to Ap_Id.

After that, execute the cfgadm(1M) command to check if the slot status is changed to "connected unconfigured".

# cfgadm -c configure BB#1-PCI#0 <Return>
# cfgadm BB#1-PCI#0 <Return>
Ap_Id                          Type        Receptacle    Occupant      Condition
BB#1-PCI#0                     etherne/hp  connected     configured    ok

3) Connect equipment to PCI cards

The extended PCI card is connected with network equipment by the cable.

4) Setup driver

The driver configuration is added by following operations.

a.
The interface name is investigated in order to configure drivers and high layer products.
Please save the result of the prtcicl (1M) command, and obtain the difference information between current result and the result taken at 1) a. Then, the driver instance number for the extended PCI card is obtained.
In the following example, since the instance number is 2, it can be determined that the interface name of the extended PCI card is "igb2."

# prtpicl -v > /tmp/prtpicl.post <Return>
# diff /tmp/prtpicl.pre /tmp/prtpicl.post | more <Return>
            :
> :devfs-path    /pci@8900/pci@4/pci@0/pci@0/network@0
> :driver-name   igb
> :binding-name  pciex8086,1521
> :bus-addr      0
> :instance      2
> :_class        network
> :name          network
            :

b.
Confirm that the interface name that is obtained from the above operation matches the one that has been added to the PCI bus slot.

c.
The configuration of each driver is added.
Please refer to each driver manual for details.

5) Connect to redundant system

Please execute following commands in order to connect the extended PCI card to the existing redundant configuration system.
The interface names described in the following procedure vary depending on the environment. For Solaris 11, the default interface name is netX (X means the instance number).

Fast switching mode
# /opt/FJSVhanet/usr/sbin/hanetnic add -n sha0 -i igb2 -f <Return>

When setting environment configuration for fast-switching mode, execute the following commands to activate the added interfaces in advance.

  • When using IPv4 address:

    • For Solaris 10

      # /usr/sbin/ifconfig igb2 plumb <Return>
      # /usr/sbin/ifconfig igb2 ipaddress netmask + broadcast + -trailers up <Return>
    • For Solaris 11

      # /usr/sbin/ipadm create-ip net2 <Return>
      # /usr/sbin/ipadm create-addr -T static -a ipaddress/netmask net2/v4 <Return>
  • When using IPv6 address:

    • For Solaris 10

      # /usr/sbin/ifconfig igb2 inet6 plumb up <Return>
    • For Solaris 11

      # /usr/sbin/ipadm create-ip net2 <Return>
      # /usr/sbin/ipadm create-addr -T addrconf net2/v6 <Return>

Note

[For Solaris 10 environment]

For Fast switching mode (IPv4), the IP address specified in this section must also be set to /etc/inet/hosts and /etc/hostname.igb2.
For Fast switching mode (IPv6), create /etc/hostname6.hme1igb1 as an empty file.
Unless these are configured, when the system reboots, the virtual interface for Fast Switching mode cannot be activated.

GS/SURE linkage mode
# /opt/FJSVhanet/usr/sbin/stphanet -n sha0 <Return>
# /opt/FJSVhanet/usr/sbin/hanetconfig create -n sha4 -m n -i IP_address -t igb2 <Return>
# /opt/FJSVhanet/usr/sbin/hanetconfig modify -n sha0 -t sha2,sha3,sha4 <Return>
# /opt/FJSVhanet/usr/sbin/strhanet -n sha0 <Return>

6) Switch the redundant path

Please switch an online communication path for extended communication path if needed.