Enabling HA for the Portal

[Keshava Murthy Jayaram]

Hi,

 

In a Docker Swarm cluster of say 3 nodes with 1 node acting as the Manager, how would one achieve HA if the manager itself goes down ?  Would you have a separate Swarm cluster which will then pick up the traffic and starts serving requests ?

 

Please let me know if some of these options are documented anywhere.

[Cornelia Postma]

For High Availability with the portal you will need to have multiple cluster. Please see

https://docops.ca.com/ca-api-developer-portal-enhanced-experience/4-2/en/install-configure-and-upgrade/scale-ca-api-portal/high-availability

for setup and more details.

[Keshava Murthy Jayaram]

Hi Conny,

 

I have gone through the docs and am still having various outstanding questions:

1. In the Deployment topology diagram, it appears as though Portal and Jarvis are installed on a single node/host. However in certain places such as the installation workflow, there is mention about Jarvis configured as an external service. Should Jarvis be configured as an External service always ?

2. In such a scenario, is Jarvis HA configuration a separate workflow. I could not find much material about how to achieve this. The Jarvis Hardware Optimizer seemed to mention a lot of nodes and components.

3. For a HA configuration of externalized DB, would we the MySQL db be configured as a cluster in an active-active mode ?

 

I am currently trying to arrive at the total number of nodes required for a production installation (that includes Portal - 3, DB -2 ? and Jarvis components - ?). Please let me know if we have some references around this

 

Also, is there some documentation around how to have an environment set up for DR purposes. Let me know if I have missed something obvious.

[Cornelia Postma]

Hi,

 

1) Jarvis needs to be installed externally and is not part of the portal cluster with 4.3.1

Any portal cluster would need to point to the same jarvis domain. (https://docops.ca.com/ca-api-developer-portal-enhanced-experience/4-3/en/install-configure-and-upgrade/install-and-configure-api-portal/integrate-jarvis-analytics and https://docops.ca.com/ca-api-developer-portal-enhanced-experience/4-3/en/install-configure-and-upgrade/configure-external-jarvis-service)

2) currently we don't have a Jarvis HA configuration, this will be part of a future release

3) for the MySQL db to be HA : the Portal will only see the MySQL database connection details as given in the portal.conf file. Please consult your DB admin for setup and configuration.

[Keshava Murthy Jayaram]

Thanks once again for the quick response Conny. A few final questions on this topic:

1. If Jarvis is not part of the portal cluster and has to be installed on a different server, as mentioned here -> Configure External Jarvis Service - CA API Developer Portal - 4.3 - CA Technologies Documentation , would it have the same software requirements in terms of OS and Docker versions ? Does it run on a container ? Is it just 1 node that is necessary to set up Jarvis ?

2. If we do not have HA as of now for Jarvis, please let me know how we should plan for failures in production ?

3. . So ideally in the Docker Swarm Cluster of Portal (3 node deployment), there is only the portal component set up in each of the worker nodes and the Swarm manager will have the Dispatcher service alone. Let me know if this understanding is not correct.

[Cornelia Postma]

1. Yes, in terms of OS and docker versions there are the same requirements. see also https://docops.ca.com/ca-api-developer-portal-enhanced-experience/4-3/en/install-configure-and-upgrade/prepare-your-environment/hardware-and-software-requirements

2. We don't have a best practise at this moment. Personally I would ensure, that the Jarvis box can be recreated from a snapshot and would keep the snapshot fairly current.

3. correct