Uploaded image for project: 'MariaDB MaxScale'
  1. MariaDB MaxScale
  2. MXS-7731

Keepalived VIP setup for multi-MaxScale deployments: peer command, generated config and scripts, pre-flight validation

    XMLWordPrintable

Details

    • New Feature
    • Status: Open (View Workflow)
    • Major
    • Resolution: Unresolved
    • None
    • None
    • None
    • None

    Description

      Problem

      The only guidance is the manual tutorial (Using Keepalived With MariaDB MaxScale). Users hand-write keepalived.conf, check and notify scripts on every node, and must choose between cooperative monitoring and passive mode themselves (combining them can leave neither MaxScale holding the monitor locks after failback). A floating VIP also cannot work across subnets, datacenters or most public clouds, and users find out only after setup.

      Goal

      Replace the manual steps with a guided flow for two or more MaxScales that reuses config sync, generates the files the tutorial describes, and stops with clear guidance when a VIP cannot work.

      Proposed flow (command names are placeholders)

      1. Requires cluster config sync; otherwise peering refuses and explains why.
      2. maxctrl create vip <name> <address/prefix> on one node (shared settings sync).
      3. maxctrl peer <host:port> from another node runs pre-flight checks.
      4. maxctrl vip plan shows generated files per node, maxctrl vip apply writes them, maxctrl show vip shows state.

      Requirements

      • Two modes: (a) VIP with cooperative monitoring, no passive toggling; (b) VIP with passive mode via a generated notify script, refused while cooperative monitoring is enabled.
      • Pre-flight: config sync and version match, same subnet, VIP valid per node, keepalived installed, VRRP reachable, cloud environment warning, SELinux commands when enforcing.
      • Generate keepalived.conf, check script (no hard-coded default credentials) and notify script; auth secret generated and distributed by peering over REST API with TLS.
      • Node-local values (interface, IP, priority) kept out of synced config, auto-detected, confirmed in the plan.
      • First release generates and validates only; MaxScale does not edit /etc/keepalived or reload keepalived.
      • Different subnets: stop and point to DNS/GSLB, load balancer, BGP or Kubernetes Service alternatives; generate nothing.

      Out of scope

      Cross-DC failover orchestration, MaxGUI wizard, changing cooperative monitoring/passive behaviour.

      Acceptance criteria

      • Two same-subnet nodes with config sync can be peered and fail over correctly in each mode.
      • Passive mode with cooperative monitoring, missing config sync, version mismatch and cross-subnet peers each fail with a clear message.
      • Tutorial updated to cover the guided flow and cross-DC alternatives.

      Attachments

        Activity

          People

            Unassigned Unassigned
            allen.herrera Allen Herrera
            Votes:
            0 Vote for this issue
            Watchers:
            1 Start watching this issue

            Dates

              Created:
              Updated:

              Git Integration

                Error rendering 'com.xiplink.jira.git.jira_git_plugin:git-issue-webpanel'. Please contact your Jira administrators.