Uploaded image for project: 'MariaDB ColumnStore'
  1. MariaDB ColumnStore
  2. MCOL-6006

CMAPI | Disable failover when shared storage not detected

    XMLWordPrintable

Details

    • New Feature
    • Status: Closed (View Workflow)
    • Major
    • Resolution: Done
    • None
    • None
    • cmapi
    • None

    Description

      Currently cmapi defaults failover to be on, as instructions "require" multinode deployments to use NFS or shared storage. However nothing checks this or blocks startup nor throws a warning, and non-HA clusters are technically possible for higher performance.

      As a user I would like cmapi to detect on startup if shared storage is used or not and disable failover if not. Having failover enabled when shared storage is not present is an issue because in the event a node goes down, i really am losing x fraction of my data and the cluster should stop working. Since each dbroot has no copy to another node, reassigning the dbroot to another node still online wont work because it'll be emtpy anyways and return tons of errors or the perception that all tables have no data (in the case of an empty dbroot1 extent map being loaded).

      Idea:
      Write a file to dbroot 1 dir, api call to another machine that checks if it can read that same file on dbroot 1, (maybe an expected hash or checksum to validate its the same file )
      Localstorage: /var/lib/columnstore/data1/x
      S3: /var/lib/columnstore/storagemanager/metadata/data1/x

      Attachments

        Issue Links

          Activity

            People

              alan.mologorsky Alan Mologorsky
              allen.herrera Allen Herrera
              Aleksei Bukhalov Aleksei Bukhalov
              Votes:
              0 Vote for this issue
              Watchers:
              4 Start watching this issue

              Dates

                Created:
                Updated:
                Resolved:

                Git Integration

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