When to use master-slave mode

Applies to Jira Data Center only.

Avoid Jira slow-downs caused by BigPicture consuming resources on all nodes. The setup described on this page can be helpful when you have issues with Jira performance that do not occur when the BigPicture app is disabled.

The master-slave mode should be treated as the last resort.

To improve performance, first upgrade BigPicture to version 8.37 or higher and set the ‘Updates processing control’ to Optimized updates processing or Limited updates processing. You can also adjust the database parameters directly.

We recommend reaching out to the Appfire Support Team for help with the configuration change.

How does it work

This setup helps minimize BigPicture's impact on Jira Data Center performance.

A load balancer redirects requests to BigPicture—changing node configuration requires changing load balancer configuration.

After the change, BigPicture will run only on the primary node. Other nodes (secondary nodes) will seamlessly redirect users' requests to the first node. As a result, only the primary node will be impacted by BigPicture’s load (processor usage, memory usage, etc.). Other nodes remain unaffected and handle Jira users' requests as usual.

Arrows show the flow of events/requests with redirections introduced in the solution.
JIRA EVENT - event sent from Jira by webhook to BP informing about change in Jira that needs to be handled in BP.
BP REQUEST - HTTP request sent by user browser to BP application.
JIRA REQUEST - HTTP request sent by user browser to Jira application.
BP AUTH/REAUTH REQUEST - HTTP authorization/reauthorization request sent by user browser to BP application.

BigPicture availability after the change

BigPictrue will still be accessible on all nodes, but requests will be handled only by the primary node. Secondary nodes will be responsible only for BigPicture menu-based actions like calculating which boxes should be shown as recently visited or menu options based on permissions. These actions don’t require a lot of resources, so the impact on secondary nodes is negligible.

You will need

competences with load balancers and access to your load balancer configuration
BigPicture in version at least 8.32, containing Master-Slave solution
testing/preprod environment to safely check configuration
decide which node will be the Master node, check its ID (defined in load balancer configuration), and ensure its hardware meets the requirements given below in this document

Steps

Prepare the load balancer configuration that is appropriate for your load balancer technology. Below, you can find an example of such a configuration. Master-Slave mode requires a load balancer to:
1. redirect requests containing /rest/softwareplant-bigpicture in path to one (Master) node,
2. DO NOT redirect requests containing any of rest/softwareplant-bigpicture/1.0/system/auth or rest/softwareplant-bigpicture/1.0/system/reauth in path,
3. delete HTTP header Cookie from redirected requests
4. delete HTTP header Set-Cookie from responses to redirected requests
Set the Master node ID using the endpoint below. After that, other nodes will act as Slave nodes.
1. enabling Master-Slave occurs by setting Master node id:
  On the node you want to set as Master node call:
  HTTP method: PUT
  path: <BASE>/rest/softwareplant-bigpicture/1.0/util/cluster/configuration
2. checking Master-Slave configuration:
  HTTP method: GET
  path: <BASE>/rest/softwareplant-bigpicture/1.0/util/cluster/configuration
3. disabling Master-Slave configuration occurs by deleting Master node id:
  HTTP method: DELETE
  path: <BASE>/rest/softwareplant-bigpicture/1.0/util/cluster/configuration
BigPicture should work in Master-Slave mode now. Check if both Jira and BigPicture work on all nodes. Jira user experience remains the same, except for better Jira performance. You can observe how Master-Slave redirects HTTP traffic, e.g., by observing HTTP headers in the browser. Depending on your load balancer solution, this could be visible by different node IDs in the BigPicture HTTP request and its response.

Load balancer configuration example

Example of load balancer configuration in standard and Master-Slave configuration for haproxy technology. If you need help with other load balancers, write to us at support@softwareplant.com https://appfire.atlassian.net/servicedesk/customer/portal/42 with the topic: “DC Master-Slave configuration example request“.

DO NOT COPY THIS AS READY SOLUTION - USE FOR REFERENCE ONLY

haproxy.cfg - Standard configuration (before change):

defaults
    (...)
frontend jira_dc
  bind 0.0.0.0:80
  default_backend jira_dc_ha
backend jira_dc_ha
  balance roundrobin
  option httpchk HEAD /dev-jira8dc/psql1/status
  cookie JSESSIONID prefix nocache
  server node01 host.docker.internal:44081 check cookie node01
  server node02 host.docker.internal:45081 check cookie node02

haproxy.cfg - Master-Slave configuration (after change) with node01 set as Master node:

defaults
    ...
frontend jira_dc
  bind 0.0.0.0:80
  default_backend jira_dc_ha
  acl is_bigpicture path_reg .*rest/softwareplant-bigpicture.*
  acl is_bigpicture_auth path_reg .*rest/softwareplant-bigpicture/1.0/system/auth.*
  acl is_bigpicture_reauth path_reg .*rest/softwareplant-bigpicture/1.0/system/reauth.*
  use_backend jira_dc_node01_ha if is_bigpicture !is_bigpicture_auth !is_bigpicture_reauth
backend jira_dc_ha
  balance roundrobin
  option httpchk HEAD /dev-jira8dc/psql1/status
  cookie JSESSIONID prefix nocache
  server node01 host.docker.internal:44081 check cookie node01
  server node02 host.docker.internal:45081 check cookie node02
backend jira_dc_node01_ha
  option httpchk HEAD /dev-jira8dc/psql1/status
  http-request del-header Cookie
  http-response del-header Set-Cookie
  server node01 host.docker.internal:44081 check

Master node hardware requirements

The server chosen to be the Master node for BigPicture should be at least as powerful as other servers.
1. If your servers differ in hardware, it is best to choose the most powerful one as Master.
If you experience any issues with BigPicture performance after the change and the lack of server resources is confirmed as the cause, it is enough to boost only the Master server.
See also BigPicture Sizing Guide | How powerful server do I need for my company?.

BigPicture Suite

Jira slow-down caused by BigPicture