Introduction
IBM PowerHA System Mirror for AIX is clustering software which gives the capability for a resource or group of resources (an application) to be automatically or manually moved to another IBM AIX® system in the event of a system failure.
Heartbeat and failure detection is performed over all interfaces available to the cluster. This could be network interfaces, Fibre Channel (FC) adapter interfaces, and the Cluster Aware AIX (CAA) repository disk.
In PowerHA 6.1 and earlier versions, heartbeat over FC adapter interfaces was not supported, and instead, a SAN-attached heartbeat disk was made available to both nodes, and this was used for heartbeat and failure detection. In PowerHA 7.1, the use of heartbeat disks is no longer supported, and configuring heartbeat over SAN is the supported method to use in place of heartbeat disks.
For this heartbeat over SAN to take place, the FC adapter in the AIX system needs to be configured to act as a target and aninitiator. In most SAN environments, an initiator device belongs to the server which is typically a host bus adapter (HBA) and a target is typically a storage device, such as a storage controller or a tape device. The IBM AIX 7.1 Information Center contains a list of supported FC adapters that can support the target mode. These adapters can be used for heartbeat over SAN.
Overview
In this article, I have provided simple examples of how to set up the SAN heartbeat in two scenarios; the first example with two AIX systems using physical I/O and the other example with two AIX logical partitions (LPARs) using Virtual I/O Server and N-Port ID Virtualization (NPIV).
In each of the examples, we have a two-node PowerHA 7.1 cluster, with one node residing on a different IBM POWER® processor-based server. This article does not cover how to configure shared storage, advanced network communications, or application controllers. This is a practical example of how to build a very simple cluster, and get the SAN heartbeat working.
Requirements
The following minimum requirements must be met to ensure that we can create the cluster and configure the SAN heartbeat:
- AIX 6.1 or preferably AIX 7.1 needs to be installed on both AIX systems, using the latest technology level and service pack.
- PowerHA 7.1 needs to be installed on both AIX systems, using the latest service pack.
- The FC adapters in the servers must support target mode, and if NPIV is in use, they must be 8 GBps adapters supporting NPIV. NPIV support is required for Scenario 2 that is explained in this article.
- If Virtual I/O Server is in use, then the VIOS code should be the latest service pack of IOS 2.2. This is required for Scenario 2 in this article.
- If NPIV is in use, then the fabric switches must have NPIV support enabled, and be on a supported level of firmware. This is required for Scenario 2 that is explained in this article.
- There must be a logical unit number (LUN) allocated to both AIX systems for use as the CAA repository disk.
- There must be a LUN allocated to both AIX systems for use as shared storage for the cluster.
Scenario 1: Two nodes using physical I/O
In this scenario, we have a very simple environment where there are two POWER processor-based systems, each with a single instance of AIX. These systems are in a PowerHA cluster and connected through redundant SAN fabrics to shared storage.
The following figure gives a high-level overview of this scenario.
Figure 1. Overview of scenario 1
SAN zoning requirements
Before the cluster can be created, SAN zoning is required. You need to configure the following two types of zones.
- Storage zones
- Heartbeat zones
To configure the zoning, first log in to each of the nodes, verify that the FC adapters are available, and capture the worldwide port number (WWPN) of each adapter port, as shown in the following example.
root@ha71_node1:/home/root# lsdev -Cc adapter |grep fcs
fcs0 Available 02-T1 8Gb PCI Express Dual Port FC Adapter
fcs1 Available 03-T1 8Gb PCI Express Dual Port FC Adapter
fcs2 Available 02-T1 8Gb PCI Express Dual Port FC Adapter
fcs3 Available 03-T1 8Gb PCI Express Dual Port FC Adapter
root@ha71_node1:/home/root# for i in `lsdev -Cc adapter |awk '{print $1}'
|grep fcs `; do print ${i} - $(lscfg -vl $i |grep Network |awk '{print $2}'
|cut -c21-50| sed 's/../&:/g;s/:$//'); done
fcs0 - 10:00:00:00:C9:CC:49:44
fcs1 - 10:00:00:00:C9:CC:49:45
fcs2 - 10:00:00:00:C9:C8:85:CC
fcs3 - 10:00:00:00:C9:C8:85:CD
root@ha71_node1:/home/root#
root@ha71_node2:/home/root# lsdev -Cc adapter |grep fcs
fcs0 Available 02-T1 8Gb PCI Express Dual Port FC Adapter
fcs1 Available 03-T1 8Gb PCI Express Dual Port FC Adapter
fcs2 Available 02-T1 8Gb PCI Express Dual Port FC Adapter
fcs3 Available 03-T1 8Gb PCI Express Dual Port FC Adapter
root@ha71_node2:/home/root# for i in `lsdev -Cc adapter |awk '{print $1}'
|grep fcs `; do print ${i} - $(lscfg -vl $i |grep Network |awk '{print $2}'
|cut -c21-50| sed 's/../&:/g;s/:$//'); donefcs0 - 10:00:00:00:C9:A9:2E:96
fcs1 - 10:00:00:00:C9:A9:2E:97
fcs2 - 10:00:00:00:C9:CC:2A:7C
fcs3 - 10:00:00:00:C9:CC:2A:7D
root@ha71_node2:/home/root#
After the WWPNs are known, zoning can be performed on the fabric switches. Zone the HBA adapters to the storage ports on the storage controller used for the shared storage, and also create zones that can be used for the heartbeat. The following diagram gives an overview of how the heartbeat zones should be created.
Figure 2. Overview of creating heartbeat zones (scenario 1)
Ensure that you zone one port from each FC adapter on the first node to another port on each FC adapter on the second node.
Device configuration in AIX
After the zoning is complete, the next step is to enable target mode on each of the adapter device in AIX. This needs to be performed on each adapter that has been used for a heartbeat zone. In the SAN zoning example, the adapters fcs0 and fcs2 on each node have been used for the SAN heartbeat zones.
For target mode to be enabled, both dyntrk (dynamic tracking) and fast_fail need to be enabled on the fscsi device, andtarget mode need to be enabled on the fcs device.
To enable target mode, perform the following steps on both nodes to zone.
root@ha71_node1:/home/root# rmdev –l fcs0 –R
fscsi0 Defined
fcs0 Defined
root@ha71_node1:/home/root# rmdev –l fcs2 –R
fscsi2 Defined
fcs2 Defined
root@ha71_node1:/home/root# chdev –l fscsi0 –a dyntrk=yes -a fc_err_recov=fast_fail
fscsi0 changed
root@ha71_node1:/home/root# chdev –l fscsi2 –a dyntrk=yes –a fc_err_recov=fast_fail
fscsi2 changed
root@ha71_node1:/home/root# chdev –l fcs0 -a tme=yes
fcs0 changed
root@ha71_node1:/home/root# chdev –l fcs2 -a tme=yes
fcs2 changed
root@ha71_node1:/home/root# cfgmgr
root@ha71_node1:/home/root#
root@ha71_node2:/home/root# rmdev –l fcs0 –R
fscsi0 Defined
fcs0 Defined
root@ha71_node2:/home/root# rmdev –l fcs2 –R
fscsi2 Defined
fcs2 Defined
root@ha71_node2:/home/root# chdev –l fscsi0 –a dyntrk=yes –a fc_err_recov=fast_fail
fscsi0 changed
root@ha71_node2:/home/root# chdev –l fscsi2 –a dyntrk=yes –a fc_err_recov=fast_fail
fscsi2 changed
root@ha71_node2:/home/root# chdev –l fcs0 -a tme=yes
fcs0 changed
root@ha71_node2:/home/root# chdev –l fcs2 -a tme=yes
fcs2 changed
root@ha71_node2:/home/root# cfgmgr
root@ha71_node2:/home/root#
If the devices are busy, make the changes with the –P option at the end of the command, and restart the server. This will cause the change to be applied at the next start of the server.
The target mode setting can be verified by checking the attributes of the fscsi devices. The following example shows how to check fscsi0 and fcs0 on one of the nodes. This should be checked on each of the fcs0 and fcs2 adapters on both nodes.
root@ha71_node1:/home/root# lsattr -El fscsi0
attach switch How this adapter is CONNECTED False
dyntrk yes Dynamic Tracking of FC Devices True
fc_err_recov fast_fail FC Fabric Event Error RECOVERY Policy True
scsi_id 0xbc0e0a Adapter SCSI ID False
sw_fc_class 3 FC Class for Fabric True
root@ha71_node1:/home/root# lsattr -El fcs0 |grep tme
tme yes Target Mode Enabled True
root@ha71_node1:/home/root#
After the target mode is enabled, we should next look for the available sfwcomm devices. These devices are used for the PowerHA error detection and heartbeat over SAN.
Check whether these devices are available on both nodes.
root@ha71_node1:/home/root# lsdev -C |grep sfwcomm
sfwcomm0 Available 02-T1-01-FF Fibre Channel Storage Framework Comm
sfwcomm1 Available 03-T1-01-FF Fibre Channel Storage Framework Comm
sfwcomm2 Available 02-T1-01-FF Fibre Channel Storage Framework Comm
sfwcomm3 Available 03-T1-01-FF Fibre Channel Storage Framework Comm
root@ha71_node1:/home/root#
root@ha71_node1:/home/root# lsdev -C |grep sfwcomm
sfwcomm0 Available 02-T1-01-FF Fibre Channel Storage Framework Comm
sfwcomm1 Available 03-T1-01-FF Fibre Channel Storage Framework Comm
sfwcomm2 Available 02-T1-01-FF Fibre Channel Storage Framework Comm
sfwcomm3 Available 03-T1-01-FF Fibre Channel Storage Framework Comm
root@ha71_node1:/home/root#
Scenario 2: Two nodes using Virtual I/O Server
In this scenario, a slightly more complex environment where there are two POWER processor-based systems, each with dual VIOS and LPARs using VIOS is used. These LPARs are in a PowerHA cluster and connected using redundant SAN fabrics to shared storage.
When using VIOS, what differs from the physical I/O scenario is that the FC ports of the Virtual I/O Server must be zoned together. There is then a private virtual LAN (VLAN) with the port VLAN ID of 3358 (3358 is the only VLAN ID that will work) used to carry the heartbeat communication over the hypervisor from the Virtual I/O Server to the client LPAR, which is our PowerHA node.
In this case, the following high-level steps are required.
- Turn on target mode on the VIOS FC adapters.
- Zone the VIOS ports together.
- Configure the private 3358 VLAN for heartbeat traffic.
- Configure the PowerHA cluster.
The following figure gives a high-level overview of this scenario.
Figure 3. Overview of Scenario 2
SAN zoning requirements
Before the cluster can be created, SAN zoning is required. You need to configure the following two types of zones.
- Storage zones
- Contains the LPAR's virtual WWPNs
- Contains the storage controller's WWPNs
- Heartbeat zones (contains the VIOS physical WWPNs)
- The VIOS on each machine should be zoned together.
- The virtual WWPNs of the client LPARs should not be zoned together.
When performing the zoning, log in to each of the VIOS (both VIOS on each managed system) and verify that the FC adapters are available, and capture the WWPN information for zoning. The following example shows how to perform this step on one VIOS.
$ lsdev -type adapter |grep fcs
fcs0 Available 02-T1 8Gb PCI Express Dual Port FC Adapter
fcs1 Available 03-T1 8Gb PCI Express Dual Port FC Adapter
fcs2 Available 02-T1 8Gb PCI Express Dual Port FC Adapter
fcs3 Available 03-T1 8Gb PCI Express Dual Port FC Adapter
$ for i in `lsdev -type adapter |awk '{print $1}' |grep fcs `;
do print ${i} - $(lsdev -dev $i -vpd
|grep Network |awk '{print $2}' |sed 's/Address.............//g'
| sed 's/../&:/g;s/:$//'); done
fcs0 - 10:00:00:00:C9:B7:65:32
fcs1 - 10:00:00:00:C9:B7:65:33
fcs2 - 10:00:00:00:C9:B7:63:60
fcs3 - 10:00:00:00:C9:B7:63:61
The virtual WWPNs also need to be captured from the client LPAR for the storage zones. The following example shows how to perform this step on both nodes.
root@ha71_node1:/home/root# lsdev -Cc adapter |grep fcs
fcs0 Available 02-T1 Virtual Fibre Channel Client Adapter
fcs1 Available 03-T1 Virtual Fibre Channel Client Adapter
fcs2 Available 02-T1 Virtual Fibre Channel Client Adapter
fcs3 Available 03-T1 Virtual Fibre Channel Client Adapter
root@ha71_node1:/home/root# for i in `lsdev -Cc adapter |awk '{print $1}'
|grep fcs `; do print ${i} - $(lscfg -vl $i |grep Network |awk '{print $2}'
|cut -c21-50| sed 's/../&:/g;s/:$//'); done
fcs0 – c0:50:76:04:f8:f6:00:40
fcs1 – c0:50:76:04:f8:f6:00:42
fcs2 – c0:50:76:04:f8:f6:00:44
fcs3 – c0:50:76:04:f8:f6:00:46
root@ha71_node1:/home/root#
root@ha71_node2:/home/root# lsdev -Cc adapter |grep fcs
fcs0 Available 02-T1 Virtual Fibre Channel Client Adapter
fcs1 Available 03-T1 Virtual Fibre Channel Client Adapter
fcs2 Available 02-T1 Virtual Fibre Channel Client Adapter
fcs3 Available 03-T1 Virtual Fibre Channel Client Adapter
root@ha71_node2:/home/root# for i in `lsdev -Cc adapter |awk '{print $1}'
|grep fcs `; do print ${i} - $(lscfg -vl $i |grep Network |awk '{print $2}'
|cut -c21-50| sed 's/../&:/g;s/:$//'); done
fcs0 – C0:50:76:04:F8:F6:00:00
fcs1 – C0:50:76:04:F8:F6:00:02
fcs2 – C0:50:76:04:F8:F6:00:04
fcs3 – C0:50:76:04:F8:F6:00:06
root@ha71_node2:/home/root#
After the WWPNs are known, zoning can be performed on the fabric switches. Zone the LPAR’s virtual WWPNs to the storage ports on the storage controller used for the shared storage, and also create zones containing the VIOS physical ports, which will be used for the heartbeat. The following figure gives an overview of how the heartbeat zones should be created.
Figure 4. Overview of creating heartbeat zones (scenario 2)
Virtual I/O Server FC adapter configuration
After the zoning is complete, the next step is to enable target mode on each of the adapter device in each VIOS. This needs to be performed on each adapter that has been used for a heartbeat zone. In the SAN zoning example, the fcs0 and fcs2 adapters on each node have been used for the SAN heartbeat zones.
For target mode to be enabled, both dyntrk (dynamic tracking) and fast_fail need to be enabled on the fscsi device, andtarget mode need to be enabled on the fcs device.
To enable target mode, perform the following steps on both VIOS on each managed system.
$ chdev -dev fscsi0 -attr dyntrk=yes fc_err_recov=fast_fail –perm
fscsi0 changed
$ chdev -dev fcs0 -attr tme=yes –perm
fcs0 changed
$ chdev -dev fscsi2 -attr dyntrk=yes fc_err_recov=fast_fail –perm
fscsi2 changed
$ chdev -dev fcs2 -attr tme=yes –perm
fcs2 changed
$ shutdown -restart
A restart of each VIOS is required, and therefore, it is strongly recommended to modify one VIOS at a time.
Virtual I/O Server network configuration
When VIOS is in use, the physical FC adapters belonging to the VIOS are zoned together. This provides connectivity between the VIOS on each managed system, however for the client LPAR (HA node) connectivity, a private VLAN must to be configured to provide this.
The VLAN ID must be 3358 for this to work. The following figure describes the virtual Ethernet setup.
Figure 5. Virtual Ethernet setup
First, log in to each of the VIOS, and add an additional VLAN to each shared Ethernet bridge adapter. This provides the VIOS connectivity to the 3358 VLAN.
The following figure shows how this additional VLAN can be added to the bridge adapter.
Figure 6. Adding the additional VLAN to the bridge adapter
Next, create a virtual Ethernet adapter on the client partition, and set the port virtual VLAN ID to be 3358. This provides the client LPAR connectivity to the 3358 VLAN.
From AIX, run the cfgmgr command and pick up the virtual Ethernet adapter.
Do not put an IP address on this interface.
Figure 7. Creating a virtual Ethernet adapter on the client partition
After this is complete, we can create our PowerHA cluster, and the SAN heartbeat is ready for use.
PowerHA cluster configuration
The first step, before creating the cluster, is to perform the following tasks:
- Edit /etc/environment and add /usr/es/sbin/cluster/utilities and /usr/es/sbin/cluster/ to the
$PATH variable.
- Populate /etc/cluster/rhosts.
- Populate /usr/es/sbin/cluster/netmon.cf.
After this is complete, the cluster can be created using smitty sysmirror or on the command line. In the following example, I have created a simple two-node cluster called ha71_cluster.
root@ha71_node1:/home/root # clmgr add cluster ha71_cluster NODES="ha71_node1 ha71_node2"
Warning: to complete this configuration, a repository disk must be defined.
Cluster Name: ha71_cluster
Cluster Connection Authentication Mode: Standard
Cluster Message Authentication Mode: None
Cluster Message Encryption: None
Use Persistent Labels for Communication: No
Repository Disk: None
Cluster IP Address:
There are 2 node(s) and 1 network(s) defined
NODE ha71_node1:
Network net_ether_01
ha71_node1 172.16.5.251
NODE ha71_node2:
Network net_ether_01
ha71_node2 172.16.5.252
No resource groups defined
Initializing..
Gathering cluster information, which may take a few minutes...
Processing...
….. etc…..
Retrieving data from available cluster nodes. This could take a few minutes.
Start data collection on node ha71_node1
Start data collection on node ha71_node2
Collector on node ha71_node1 completed
Collector on node ha71_node2 completed
Data collection complete
Completed 10 percent of the verification checks
Completed 20 percent of the verification checks
Completed 30 percent of the verification checks
Completed 40 percent of the verification checks
Completed 50 percent of the verification checks
Completed 60 percent of the verification checks
Completed 70 percent of the verification checks
Completed 80 percent of the verification checks
Completed 90 percent of the verification checks
Completed 100 percent of the verification checks
IP Network Discovery completed normally
Current cluster configuration:
Discovering Volume Group Configuration
root@ha71_node1:/home/root #
After creating the cluster definition, the next step is to check whether there is a free disk on each node, so that we can configure the CAA repository.
root@ha71_node1:/home/root# lsdev –Cc disk
hdisk0 Available 00-00-01 IBM MPIO FC 2107
hdisk1 Available 00-00-01 IBM MPIO FC 2107
root@ha71_node1:/home/root# lspv
hdisk0 000966fa5e41e427 rootvg active
hdisk1 000966fa08520349 None
root@ha71_node1:/home/root#
root@ha71_node2:/home/root# lsdev –Cc disk
hdisk0 Available 00-00-01 IBM MPIO FC 2107
hdisk1 Available 00-00-01 IBM MPIO FC 2107
root@ha71_node2:/home/root# lspv
hdisk0 000966fa46c8abcb rootvg active
hdisk1 000966fa08520349 None
root@ha71_node2:/home/root#
From the above example, it is clear that hdisk1 is a free disk on each node. So, this can be used for the repository. Next, modify the cluster definition to include the cluster repository disk. Our free disk on both nodes is hdisk1.
This can be performed using smitty hacmp or on the command line. The following example shows how to perform this step on the command line.
root@ha71_node1:/home/root # clmgr modify cluster ha71_cluster REPOSITORY=hdisk1
Cluster Name: ha71_cluster
Cluster Connection Authentication Mode: Standard
Cluster Message Authentication Mode: None
Cluster Message Encryption: None
Use Persistent Labels for Communication: No
Repository Disk: hdisk1
Cluster IP Address:
There are 2 node(s) and 1 network(s) defined
NODE ha71_node1:
Network net_ether_01
ha71_node1 172.16.5.251
NODE ha71_node2:
Network net_ether_01
ha71_node2 172.16.5.252
No resource groups defined
Current cluster configuration:
root@ha71_node1:/home/root #
The next step is to verify and synchronize the cluster configuration. This can be performed using smitty hacmp or on the command line. The following example shows how to synchronize the cluster topology and resources on the command line.
root@ha71_node1:/home/root # cldare -rt
Timer object autoclverify already exists
Verification to be performed on the following:
Cluster Topology
Cluster Resources
Retrieving data from available cluster nodes. This could take a few minutes.
Start data collection on node ha71_node1
Start data collection on node ha71_node2
Collector on node ha71_node2 completed
Collector on node ha71_node1 completed
Data collection complete
Verifying Cluster Topology...
Completed 10 percent of the verification checks
WARNING: Multiple communication interfaces are recommended for networks that
use IP aliasing in order to prevent the communication interface from
becoming a single point of failure. There are fewer than the recommended
number of communication interfaces defined on the following node(s) for
the given network(s):
Node: Network:
---------------------------------- ----------------------------------
ha71_node1 net_ether_01
ha71_node2 net_ether_01
Completed 20 percent of the verification checks
Completed 30 percent of the verification checks
Saving existing /var/hacmp/clverify/ver_mping/ver_mping.log to
/var/hacmp/clverify/ver_mping/ver_mping.log.bak
Verifying clcomd communication, please be patient.
Verifying multicast communication with mping.
Verifying Cluster Resources...
Completed 40 percent of the verification checks
Completed 50 percent of the verification checks
Completed 60 percent of the verification checks
Completed 70 percent of the verification checks
Completed 80 percent of the verification checks
Completed 90 percent of the verification checks
Completed 100 percent of the verification checks
… etc…
Committing any changes, as required, to all available nodes...
Adding any necessary PowerHA SystemMirror entries to
/etc/inittab and /etc/rc.net for IPAT on node ha71_node1.
Adding any necessary PowerHA SystemMirror entries
to /etc/inittab and /etc/rc.net for IPAT on node ha71_node2.
Verification has completed normally.
root@ha71_node1:/home/root #
Now that a basic cluster has been configured, the last step is to verify that the SAN heartbeat is up.
The lscluster –i command displays the cluster interfaces and their status. The sfwcom (Storage Framework Communication) interface is the SAN heartbeat.
In the following example, we can check this from one of the nodes to ensure that the SAN heartbeat is up. This is good news!
root@ha71_node1:/home/root # lscluster -i sfwcom
Network/Storage Interface Query
Cluster Name: ha71_cluster
Cluster uuid: 7ed966a0-f28e-11e1-b39b-62d58cd52c04
Number of nodes reporting = 2
Number of nodes expected = 2
Node ha71_node1
Node uuid = 7ecf4e5e-f28e-11e1-b39b-62d58cd52c04
Number of interfaces discovered = 3
Interface number 3 sfwcom
ifnet type = 0 ndd type = 304
Mac address length = 0
Mac address = 0.0.0.0.0.0
Smoothed rrt across interface = 0
Mean Deviation in network rrt across interface = 0
Probe interval for interface = 100 ms
ifnet flags for interface = 0x0
ndd flags for interface = 0x9
Interface state UP
root@ha71_node1:/home/root #
The remaining steps for cluster configuration, such as configuring shared storage, mirror pools, file collections, application controllers, monitors, and so on are not covered in this article.
![Join Us on Facebook](https://2.bp.blogspot.com/-7_3aMhrvLq4/UTtUB9H5ZyI/AAAAAAAABcM/JjZGcP7r3_I/s1600/fbb.png")
![Follow us on Twitter](https://2.bp.blogspot.com/-VXfOkCT43uI/UTtUEItlzyI/AAAAAAAABck/cvROEjW1Vx8/s1600/tww.png")