Jump to: navigation, search

Traditional HA Failure Scenarios


When an application failure happens, two types of switchover scripts are invoked: Virtual IP (VIP) control, and application control. You must add VIP control scripts on the host where the voicemail components are installed, and application control scripts on the host where SCS is running. SCS executes a script based on alarm reactions to execute the VIP and Application control scripts. For details about the events, scripts and third-party applications, see: Traditional HA Events and Scripts. The following are failure scenarios for Traditional HA, where a Primary GSVM SIP Server or Primary GSVM Server failure results in a coordinated switchover of the components, from primary to backup server, using the MLCMD utility, which is installed with SCS. The switchover for GVP components is defined by the individual HA deployment modes of GVP and Premise SIP Server.

Primary Host Failure

The following items describe the switchover mechanism using the MLCMD utility when the primary VM host components fail while running in primary mode.

VM SIP Server-1 Fails

  • The VM SIP Server-2 generates a 4563 event.
  • Alarm reaction scripts are triggered to react to the 4563 event from the VM SIP Server-2.
  • VIP takeover scripts are linked to only the primary and backup VM SIP Servers on the VM host.

MLCMD Reaction:

  1. The VM Server-2 is switched to primary mode.
  2. VIP is enabled on VM Host-2.
  3. VIP is disabled on VM Host-1.
  4. There is no impact on the GVP Host.

VM Server-1 Fails

  • VM Server-2 generates a 4563 event.
  • Alarm reaction scripts are created to react to the 4563 event from VM Server-2.
  • VIP takeover scripts are linked to only primary and backup VM SIP servers.
  • The 4563 event from VM Server-2 switches VM SIP Server-2 to primary mode in the VM Host.
  • The 4563 event from VM SIP Server-2 executes the VIP control scripts in the VM Host.

MLCMD Reaction:

  1. The VM SIP Server-2 is switched to primary mode.
  2. The 4563 event from VM SIP Server-2 executes the VIP control scripts.
  3. There is no impact on the GVP Host.

Resource Manager-1 Fails

When Resource Manager-1 fails, the actions are defined by the HA deployment for Resource Manager-1. Refer to the deployment guide of the Resource Manager.

Media Control Platform-1 Fails

When Media Control Platform-1 fails, the actions are defined by the HA deployment for Media Control Platform. Refer to the deployment guide of the Media Control Platform.

Entire VM Host Fails

  • Any of the following events automatically triggers the switchover, because the alarm reaction is configured for each component individually.
  • The 4563 event from VM SIPServer-2 or VM Server-2.
  • VM SIP Server-2 & VM Server-2 automatically become primary.
Note: In this scenario there are chances of getting an IP conflict problem. See Known Issues and Recommendations for more information.

MLCMD Reaction:

  1. VIP is enabled in VM Host-2.
  2. VIP will be disabled in VM Host-1.

Backup Host Failure

If any component on the backup VM host fails while running in Primary Mode, the event 4563 is generated from the components and alarm reaction scripts are executed in the reverse direction to switch the primary VM host components to primary mode.

Switchover Cases

During manual switchover scenarios, the 4563 event is generated from the VM Server-2 and the reaction scripts are the same as when the VM Server-1 fails. For example, an administrator switches the VM Server-2 to primary mode using the SCS switchover option. When an administrator switches the VM SIP Server-2 to primary mode using the SCS switchover option, the 4563 event is generated from the VM SIP Server-2 and the reaction scripts are the same as those for the VM SIP Server-1 failure.

Key Notes:

  • Failure of the components on the GVP host are handled using their native failure mechanism.
  • Failure of the GVP host components do not affect the current state of the VM host components.
  • When an application fails while running in backup mode, no alarm reaction scripts are triggered. The application does not have HA until the backup server is operational.
  • If the Premise SIP server goes down, no alarm reaction scripts are triggered.
  • MLCMD must use the application DBID. In Management Framework release 8.1.1, MLCMD can use the application name.
  • VIP takeover scripts are only linked to the event for only the VM SIP Server-1 and VM SIP Server-2. This prevents duplicate execution of the takeover scripts. If any application fails on a VM host, all applications are switched. Alarm reaction scripts are triggered for the events from different applications for the single switchover scenario.
  • VIP takeover scripts are connected to the SIP Server so that the VM SIP Server port is bonded to the VIP immediately when it becomes primary. If the VIP is not enabled immediately after SIP server becomes primary, the SIP server is in an unavailable state.
For example, when VM SIP Server fails, if the VIP scripts are connected to the VM Server, VM SIP server-2 is in an unavailable state until the 4563 event from the VM SIP Server-2 triggers the switchover of VM Server-2 and the 4563 event from VM Server-2 triggers the VIP enable script. VM SIP server-2 remains in an unavailable state until the Virtual IP is enabled in the host that becomes primary.
Note: This issue does not apply to the other applications, because the port and VIP bonding of the other applications do not happen at the switchover, when the applications become primary.


Feedback

Comment on this article:

blog comments powered by Disqus
This page was last modified on May 16, 2014, at 08:20.