Sun Fire 15K Server
 
Sun Fire E20K Server
 
Sun Fire E25K Server
 
Sun Fire 12K Server
 

PLA-Support>Sun Systems>SPARC>Enterprise>SN-SPARC: SF-Exxk
 
_Old GCS Categories>Sun Microsystems>Servers>High-End Servers
 

Applies to:

Goal

Explanation of the Main System Controller's failover to the Spare when all the Community Network (C1) interfaces fail.

Solution

Description
A failure of all configured C1 (Public) network interfaces will cause the SMS Main System Controller (SC) role to failover to the Spare and the outgoing Main to reboot.

This is a normal behavior which is designed for the failover mechanism on the SC's on a Sun Fire[TM] 12K,15K,E20K,E25K.

Steps to Follow
On a C1 network configured using IPMP on SCs, if the IPMP "test" network interface fails its health check (Typically a ping test to a router or a multicast to the network) IPMP will try to fail over the logical IP to another member of the IPMP group.

If the IPMP group is completely unavailable then the Solaris[TM] "policing" action takes over. At that point the Failover Management Daemon (fomd) may have to take action and possibly then force a failover to the Spare SC.

The following message will be logged into the Main SC's /var/adm/messages before it reboots :

The $SMSVAR/adm/platform/messages file will have the following error messages logged:

There may be various reasons why the both the interfaces on the C1 network fails. Typical scenarios may include but are not restricted to:

• Routers or bridges not responding or the end device connected to the C1 network ports are not responding.
• IPMP not configured properly for the C1 network. Document 2028757.1
• Worst case , but least probable - both the network interfaces are malfunctioning.

The showfailover -v command on the Main SC will give you the status of the C1 network:

The ifconfig -a command on the Main SC will give you details about the C1 network interfaces:

Internal Section

Some internals on how the Community network related test works on the Main SC:

XnetTest is a thread which belongs to the fomd (Failover Management Daemon) daemon. The function of the XNetTest thread is to test the external network.
This test checks for the presence of incoming packets on each of the external network interface adapters. This information is retrieved using the Solaris kstat routines.
For each of the configured external network interfaces, the test retrieves the current rxpkt (Received Packet) count, waits for two seconds, and then retrieves the rxpkt count again.
The two counts are then compared. If they differ then the given network interface is considered to have passed the test.
If the two counts are equal, the test proceeds to attempt to generate incoming packets on the given network interface by pinging the multicast router address (224.0.0.2), the multicast broadcast address (224.0.0.1) and finally the subnet broadcast address (x.x.x.255). If all of these pings fail to generate any incoming packets on the given network interface, that interface is considered to have failed the test.

If there is at least one good interface on a given SC, the XNetTest posts a TEST_PASSED to the FM, otherwise it posts TEST_FAILED. This will lead to the SC failover if it is the Main SC. If routing over the I2 network is allowed then a XNetTest failover on the main SC will never trigger a failover.
A XNetTest failure on the spare SC will always disable the failover mechanism because there is no advantage in failing over to the spare SC if the spare is unreachable from the external network.

References:
Document 1018896.1: Sun Fire[TM] 12K/15K/E20K/E25K: Management Networks (MAN)
Document 1361846.1: Sun SPARC Sun Fire 12K/15K/E20K/E25K (Starcat) Product Pages
Starcat System Controller Failover Document

Keywords: SC, C1 network, Community network, 12K, 15K, 20K, 25K, SMS, fomd, MAN Network

Previously Published As 82854

References

This solution has no attachment