The GigaVUE‑FM High Availability (HA) feature supports a highly available fabric management environment with minimal interruption. The GigaVUE‑FM HA architecture consists of a minimum of three GigaVUE‑FM instances that run together as a highly available group. The highly available group provides protection from failure of any one of the members in the group. The communication between the GigaVUE‑FM instances in the highly available group can be encrypted using Pre-shared Secret Key or Certificate based Authentication.

The following figure shows the high-level architecture of the GigaVUE‑FM HA feature.

Starting in software version 5.14.00, you can add additional GigaVUE-FM instances to the high availability group. Refer to the Dynamic Addition of GigaVUE-FM Instances to HA Group section for details.

To configure the GigaVUE‑FM HA feature, you must have access to at least three authenticated GigaVUE‑FM instances that reside on a trusted network. All GigaVUE‑FM instances must run the same software version. The interfaces in the GigaVUE‑FM instances must be up and must be assigned IPv4/IPv6 addresses. You can also choose to use DNS host names.

Note:  You can configure the GigaVUE‑FM HA feature only if you have administrative privileges. You must have Prime license installed for forming the GigaVUE-FM HA group.

The GigaVUE‑FM instances are not required to be in the same subnet, but still must be able to communicate with each other.

In addition, ensure that the instances have unique host names. You must be able to ping a GigaVUE‑FM instance from the other instances using the hostname or the IP address.

To add a GigaVUE‑FM instance to a GigaVUE‑FM HA group:

You must install a Prime license on the active GigaVUE‑FM instance to configure a High Availability group.

If the Prime license expires or if you accidentally delete the license, the existing configurations will still be present in the GigaVUE-FMs that are part of the HA group, but you will not be able to perform any new configurations. Moreover, if you disassemble the HA group, you cannot reconfigure the HA group without installing a valid Prime license. You can re-use the same Prime license to form the HA group again.

The GigaVUE‑FM HA feature is supported on the following platforms:

Keep in mind the following rules and notes when you configure the GigaVUE‑FM HA feature:

Until software version 5.13.00, GigaVUE-FM HA architecture is based on 2N+1 redundancy model (N = 1) with one active and two standby instances providing resiliency and fault tolerance. With horizontal scaling, the services and application requests are distributed to standby GigaVUE-FMs. As the HA group was limited to only three GigaVUE-FM instances, there was a need for higher computing and memory intensive resources in the HA group because of the following use cases:

• Longer data retention period (up to 13 months).
• Better search performance with more horizontal distribution of OpenSearch for features such as Fabric Health Analytics (FHA) and Topology Visualization.
• Better indexing performance with the number of instances collecting various statistics on the configured components.

Therefore, the number of GigaVUE-FM instances in the HA group had to be increased for scalable performance of the various features in GigaVUE-FM.1 GigaVUE-FM software version 5.13.00 is scaled to manage 3000 nodes. However, with use cases such as Topology Visualization, Fabric Health Analytics that rely on Elastic Search database, Elastic search has become a bottleneck for efficient performance. Therefore, it is required to increase the number of ES instances for enhanced performance and scalability of the HA group. Starting in software version 5.14.00, GigaVUE-FM allows you to add more than three GigaVUE-FM instances to the HA group. The hardware requirements and resource allocation process for the additional instances are the same as that of the active and standby instances. The additional instances can be configured either as supporting instances or as active-eligible instances:

Supporting instances are GigaVUE-FMs that run minimal services, thereby providing more resources to OpenSearch services. Supporting instances:

• do not distribute application requests (all application services are turned off by default) nor run any distributed services.
• offer more scalability.
• cannot become candidates for active instance election

Active-Eligible instances are GigaVUE-FMs that distribute application requests and run distributed services. Active-eligible instances:

• Add more value to fault-tolerant and high available system.
• Can be part of active instance election.

Refer to the following diagram for the roles in the HA group.

Note:  When you add a GigaVUE-FM instance as either a supporting instance or a standby instance, the current configuration of that GigaVUE-FM instance will be removed once it is added to the existing GigaVUE-FM HA cluster.

The following table compares GigaVUE-FM HA in software version 5.13.00 and less with GigaVUE-FM HA software version 5.14.00:

Keep in mind the following recommendations and notes when working with the HA group.

GigaVUE-FM cannot support more than 7 active-eligible nodes (this is due to the underlying constraint in MongoDB).

Shrinking the size of the HA group:

• When reducing the size of a HA group that has both active-eligible and supporting instances, it is always recommended to remove the supporting instances first so that the active-eligible instances remain intact in the HA group. If active-eligible instances are removed first from the HA group, then the HA group will become unstable. This, however, depends on the number of active-eligible and supporting instances in the HA group. Consider a GigaVUE-FM HA group with 7 instances. The number of active and supporting instances in a HA group may be as follows:

HA group 1	With 3 active instances and 4 supporting instances	The recommendation is specifically applicable for HA group 1 in which there are only 3 active eligible instances in the group. Removing the active-eligible instances may cause the HA group to become unstable.
HA group 2	With 5 active instances and 2 supporting instances.	The recommendation is not applicable for HA group 2 as removing the active-eligible instances will not impact the stability of the HA group.

For a HA group to be stable, it must have a minimum of two active-eligible instances in the group.

• When reducing the size of HA group from 7 instance to 5 instance or 3 instance, the quorum is re-computed and adjusted accordingly. This is to ensure that the HA group remains intact with the available set of instances.
• When reducing the size of the HA cluster, during removal of an instance make sure to give time for the OpenSearch to rebalance and return to a complete (100%) healthy state before removing the next instance. If you remove the instances subsequently one after the other, the health state of the HA group will turn red as there will be no time for shard rebalancing amongst the available instances.
  Login to GigaVUE-FM CLI as a root user and execute the following command to check the ES cluster health:
  

curl -XGET "http://localhost:9200/_cluster/health?pretty"

Supporting GigaVUE-FM Instances:

• When a supporting instance is added to the HA group, its system information will not be available immediately and will be available only after a minute.
• When the HA group is upgraded, application services of the supporting instances will be turned on automatically for the upgrade and once GigaVUE‑FM HA upgrade is completed it will be turned off automatically. Application services will be turned off automatically irrespective of the upgrade completion status, that is, either success or failure.

When you login to any of the GigaVUE-FM instances of the HA group, the dashboard page appears. Use the drop-down option in the header of GigaVUE‑FM, which is available on specific pages, to view the details pertaining to that GigaVUE‑FM. For example, the following pages in the GigaVUE‑FM GUI have drop-down option in the header:

• FM Health
• IP Resolver
• NTP Server for Clock Synchronization

GigaVUE‑FM instances participating in the HA group act as Load Balancers allowing better distribution of load within the GigaVUE-FM instances. GigaVUE-FM load balancer functionality provides the following capabilities:

• Seamless access to the GigaVUE-FM Dashboard page: Accessing any GigaVUE-FM GUI always takes you to the GigaVUE-FM dashboard page after successful login. This provides a cluster view of the GigaVUE-FM GUI rather than individual views for the active and standby instances.
• Ability to access the available GigaVUE-FM GUI even during a failover: However, there will be an impact to the write operation, until a new active instance takes over. For example, when you create, update or delete any of the resources in GigaVUE-FM such as maps, GigaSMART groups, tags, etc. during failover, then the operation will fail with the following error message: Unable to Connect to Server.
• Enhanced distribution of load across the members of the HA group: This allows faster response to the HTTP GET requests.
• Ability to perform backup/restore/upgrade operation from any of the GigaVUE-FM instances.

GigaVUE-FM Load Balancer Replaces Get Distribution Support

The GigaVUE-FM Load Balancer functionality replaces the GetDistribution support provided in software version 5.12.00.

Note:   You cannot disable the Load Balancer functionality.

The following behavioral changes are observed:

• After creating, updating, or deleting resources in GigaVUE-FM, the GigaVUE‑FM GUI will be updated immediately. However, if there is latency between the GigaVUE-FM instances, the GUI will not be updated immediately. The subsequent refresh will display the updated data.
• GetDistribution will be impacted if DNS resolution of GigaVUE-FM instances fail:
  • If active GigaVUE-FM instance DNS is unresolvable, then the load will be distributed between the two standby instances, with an impact to the write operation.
  • If standby GigaVUE-FM DNS is unresolvable, then the load on the active instance increases, thereby increasing the CPU and Memory Usage.
• If any GigaVUE-FM instance is down in the HA group, and if a API request is forwarded to that instance due to load balancing, then for a few seconds GigaVUE-FM GUI will not load any data. The subsequent refresh or reload will display the updated data.

Note:  Consider three GigaVUE-FM instances FM-A, FM-B, and FM- C and that you are logged in to the HA group with the IP address or DNS name of FM-A. When working with the GigaVUE-FM GUI, if FM-B goes down, then the GUI will not load any response. This is because the load balancer (FM-A in this case) has not learned about the unreachability of FM-B instance, which will happen only during the next periodic health checkup.

• The rate at which the active GigaVUE-FM instance handles the HTTP Get request and its processing will be high in case of HA upgrade when the standby GigaVUE-FM instances are upgrading and that forces active GigaVUE-FM to handle all the HTTP Get requests. In this case, CPU, Memory Usage will be slightly higher than the normal.

Important Recommendations

• For seamless access to the HA group, configure a DNS name that resolves to all the GigaVUE-FM instances participating in the cluster.
• For better user experience, the backup/restore operation must be carried out from an active GigaVUE-FM instance.
  • If you perform a restore operation from active instance, you will be notified about the restore operational status before the GigaVUE-FM goes down.
  • If you perform the same restore operation from a standby GigaVUE-FM instance, GigaVUE-FM GUI will go down without any notification about the restore operation to the users.
• If GigaVUE-FM HA upgrade fails and if the upgrade is completed manually, you must login to the GigaVUE-FM CLI as root user and run the following command. To login as root user, login into shell with admin credentials and type "sudo su -"

curl -XPOST "localhost:4466/fmcs/updateLoadBalancer?pretty"

Response :

{"operation":"success"}

Note:  Contact Gigamon Technical Support if you do not see this response.

To remove or replace a standby GigaVUE‑FM instance from the GigaVUE‑FM HA group follow the below listed steps:

12.

Go to High Availability.

Note:  The status of the GigaVUE‑FM HA group changes from Healthy to At Risk. You will not be allowed to remove the other standby GigaVUE‑FM instance after the HA status changes to At Risk. To remove the standby node, if there is any network latency in GigaVUE-FM and the node is not removed correctly, use the command fmcs remove force in the GigaVUE-FM CLI of the standby node.

To completely disassemble the GigaVUE‑FM HA group:

1.

Go to High Availability.

Note:  You cannot remove an active GigaVUE‑FM instance or disassemble the GigaVUE‑FM HA group by logging in from a standby GigaVUE‑FM instance.

Note:  Executing the above steps disassembles the GigaVUE‑FM HA group completely. The database of the individual GigaVUE-FM HA group members (active, standby, and supporting instances) will be erased and reinitiated with the default database. If you had performed orchestrated configuration flows (such as the Flexible Inline Arrangement or Intent Based Orchestration) you must manually remove and recreate these if they do not backup the GigaVUE‑FM database before choosing Delete HA Group.

If you cannot access the active GigaVUE-FM instance's GUI, use the following CLI command in all the GigaVUE-FM instances to disassemble the HA group:
/opt/fmcs/bin/fmcs leave force
Before disassembling the instances, take a backup of the configuration. After the GigaVUE‑FM instances are disassembled from the HA group, the configurations present in the instances will be completely removed.

The GigaVUE‑FM HA state depends on the status of the three GigaVUE‑FM instances. The following table lists the various states of the GigaVUE‑FM HA group. You can view the HA group state from Administration > High Availability in the GigaVUE‑FM GUI.

In case of FMHA going to suspended mode from At-Risk mode, the reachability status of node which went down (due to which HA went into suspended mode) won't be audited as writes are not accepted if there is no active/leader instance in the DB cluster

The active GigaVUE‑FM instance in the high availability group may fail at times resulting in one of the standby instances to take over and become the active instance. This process is called failover.

The following table provides the reasons for failover:

The High Availability page ( Go to High Availability) displays the current state of the GigaVUE‑FM HA group. When a failover occurs, the HA group state changes in the GUI.

The following table lists the GUI changes for the various scenarios:

Use the following table to troubleshoot issues that you might encounter while working with the HA feature.

Orchestrated upgrade of GigaVUE-FM instances in a High Availability group is similar to upgrading a standalone GigaVUE-FM instance. You can upgrade using an image that is located on an external image server, or you can use GigaVUE-FM as the image server. Refer to the Upgrade GigaVUE-FM section in the GigaVUE-FM Installation and Upgrade Guide for more details.

Note:  

• Orchestrated Upgrade of GigaVUE-FM instances in a HA group is supported from software version 5.10.01. For GigaVUE-FM software version 5.10.00 and above, it is recommended only to use the orchestrated upgrade procedure to upgrade the GigaVUE-FM instances in a HA group.
• When upgrading from any previous version to 6.10 or above, the GUI may appear stuck in the reboot state, as a new certificate is generated using the GigaVUE-FM PKI service. Reload the page to update the status. This issue will not occur in subsequent upgrades.

Prerequisites

Before upgrading the GigaVUE-FM instances in a High Availability group, ensure the following:

• The High Availability group must be in a healthy state.
• The latency between the GigaVUE-FM instances must be less than 100ms.
• The config disk space allocated to GigaVUE-FM must have a maximum sustained transfer rate of above 100MB/s. A low disk rate impacts both file sync and installation.

Upgrading GigaVUE-FM Instances in HA Group

To upgrade the GigaVUE-FM instances in a High Availability group from the GUI, click the Upgrade option from the User icon. Always trigger the upgrade from the active GigaVUE-FM instance.

Note:  Do not run any scripts (example: scripts for fetching polling statistics of maps) or queries (example: database or stats queries) prior to or during the upgrade operation as this will drain GigaVUE-FM's memory and CPU resources and impact the upgrade operation.

The following is the sequence of events that occur in the background:

1. Active GigaVUE-FM Instance: Software image download process is triggered and the image is downloaded.
2. Active GigaVUE-FM instance: Syncs and copies the downloaded image with one of the standby GigaVUE-FM instances - the first standby GigaVUE-FM instance.
3. First Standby Instance: Image is synced, and the standby instance will get upgraded first and rebooted.
4. Second Standby Instance: Image is then synced by the second standby instance and the second standby instance will get upgraded and rebooted.
5. Active GigaVUE-FM Instance: Once the standby instances are upgraded, the active GigaVUE-FM will start to upgrade and will be rebooted.
6. A new active GigaVUE-FM instance will be elected while the active GigaVUE-FM reboots.

Note:   As orchestrated upgrade is a background process, device management tasks will be carried out seamlessly. The overall time consumed for the upgrade process is around 60 minutes and the total management loss time during orchestrated upgrade is around 1 minute. This is the time required to elect the new active GigaVUE-FM instance when the current active GigaVUE-FM instance reboots post upgrade.

In a GigaVUE-FM high availability environment, you can perform the GigaVUE-FM and device configurations only from the active GigaVUE-FM instance. Standby instances provide minimal configuration options. In case of failover, it is important to be aware of the IP addresses of the three GigaVUE-FMs and also the DNS host names of the GigaVUE‑FM instances so that you can access the active GigaVUE‑FM instance.

To overcome this restriction and still access the GigaVUE‑FM instances, you can use one of the following options:

• Use Load Balancer
• Assign a DNS Name for the GigaVUE‑FM instances

Use Load Balancer

Load balancer distributes traffic across a number of servers. Integrating a load balancer with GigaVUE‑FM, forwards the traffic to the active GigaVUE‑FM instance. Load balancer performs healthcheck of GigaVUE‑FM and forwards the traffic destined to the external GigaVUE‑FM IP address and thereby to the GigaVUE-FM high availability group.

To use load balancer for forwarding the traffic, you must ensure to do the following:

• Deploy the load balancer.
• Configure the load balancer to access the active instance in the GigaVUE-FM High Availability group. This is accomplished using the following GET API endpoint exposed by GigaVUE-FM:

https://<FM-IP>/api/v1.3/fmHa/status

Based on the returned response codes, the load balancer can be configured in such a way that healthy servers (active instance) are those that return 200 as response code when the endpoint API is queried. The remaining servers are considered as unhealthy servers (standby instances). With these configurations, the requests are forwarded to active GigaVUE-FM always.

Sample Response for GET https://<FM-IP>/api/v1.3/fmHa/status

If FM Role == active/standalone,

return 200 with Payload {"role" : "active" } or {"role" : "standalone" }

Else if FM Role == Standby

return 202 with Payload {"role" : "Standby" }

Else if FM Role == support, suspended, unknown

return 400 with Payload for 400 {"role" : "Unknown" }

Note : 5xx can be thrown if the API Gateway on FM is not available/down

 

• Configure the load balancer with external GigaVUE-FM IP address as the DNS IP, for example, myfm.com.

Note:  If you add or remove the nodes in the GigaVUE-FM HA group, you must ensure to update the corresponding IP addresses of the nodes in the load balancer.

Assign DNS Name for the GigaVUE-FM IP 

You can assign a DNS name to the three GigaVUE‑FM IP addresses, which helps to access the GigaVUE-FM instance in case of failures. Consider a DNS name, myfm.com that has the IP addresses of the three GigaVUE‑FM instances of the high availability group.

If you type the DNS name of the GigaVUE-FM, myfm.com, the DNS server returns the IP addresses of the three GigaVUE‑FM instances, and:

• the Dashboard page of the active GigaVUE-FM instance appears, or
• the High Availability page of the standby GigaVUE-FM instances may appear if the active instance is not reachable, as in the case of a failover. From the High Availability page of the standby instance, you can navigate to the GigaVUE‑FM active instance.

This ensures that the GigaVUE-FM High Availability group is always accessible.