Even though there is no error in network devices, the following message is output and HUB monitoring ends abnormally.
hanet: ERROR: 87000: polling status changed: primary polling failed. (hme0,target=192.13.71.20) hanet: ERROR: 87100: polling status changed: secondary polling failed. (hme1,target=192.13.71.21) |
In NIC switching mode, occasionally it takes time to establish a data link at Ethernet level following activation of an interface. Even though activated an interface, it is not possible to communicate immediately. Generally it becomes possible to communicate in dozens of seconds after activated, but some HUBs to connect take more than one minute, and occasionally ping monitoring fails and switching occurs.
In such a case, extend the time to wait for linking up (default value: 60 seconds) by a hanetpoll on command. Also when HUB to use is set to use STP (Spanning Tree Protocol), occasionally takes long time to become possible to communicate. Extend the time to wait for linking up if necessary. On the HUB where STP is running, possible next connection could take twice as the transfer delay time(normally 30 sec) after linked up. Standard link up latency of operating STP can be derived from the equation below. For verifying STP transfer delay time, see the manual of HUB your using.
link up latency > STP transfer delay time * 2 + monitoring period * number of monitoring |
Note
To operate ping monitoring over the system that runs firewall, configure the firewall so that ping can pass through the firewall. Otherwise, it fails to operate ping monitoring.
The firewall settings must be the same for both of the primary and secondary interfaces.
Takes time to execute an operation command of a Redundant Line Control function.
Takes time to activate a userApplication or to switch nodes at the cluster operation.
When a host name or an IP address specified in the information of a virtual interface, the monitor-to information, etc. is not described in /etc/inet/hosts file, or when "files" are not specified at the top in an address solution of /etc/nsswitch.conf, occasionally it takes time to process an internally executed name-address conversion. Therefore, it takes time to execute a command, or for the cluster state to change. Check that all IP addresses and host names to use in a Redundant Line Control function are described in /etc/inet/hosts, and that /etc/inet/hosts is referred first at name-address conversion.
Even though TCP communication by a virtual IP is executed in GS/SURE linkage mode, the number of the connections is not shown when displayed how the connection is divided using a dsphanet command.
When dividing TCP connection in GS/SURE linkage mode, necessary to define the information of the other system with a hanetobserv command. Any protocol other than TCP is not divided. UDP and ICMP are sent according to the route information.
