Uploaded image for project: 'Kudu'
  1. Kudu
  2. KUDU-2135

Persist disk health state to disk

    XMLWordPrintableJSON

    Details

    • Type: Improvement
    • Status: Open
    • Priority: Major
    • Resolution: Unresolved
    • Affects Version/s: None
    • Fix Version/s: None
    • Component/s: fs
    • Labels:
      None

      Description

      When a tablet server disk fails, it is marked FAILED in memory and not touched during the lifetime of the tablet server. The next time the tablet server is started, however, if the disk happens to start up successfully, it will be used as is.

      This may be risky, as the disk may be corrupted, or may be more prone to runtime failures. Additionally, when we begin striping metadata or WALs, we may end up with multiple of them for a single tablet (e.g. one that was on the failed disk, and another if the tablet was reassigned to the same tablet server). As such, the contents of the previously failed disk should not be used. It is thus necessary to persist the health of a disk to ensure unhealthy disks are not used.

        Attachments

          Issue Links

            Activity

              People

              • Assignee:
                andrew.wong Andrew Wong
                Reporter:
                andrew.wong Andrew Wong
              • Votes:
                0 Vote for this issue
                Watchers:
                1 Start watching this issue

                Dates

                • Created:
                  Updated: