mdadm checkraid cronjob runs at the same time

Bug #1815201 reported by Wouter van Bommel on 2019-02-08
14
This bug affects 1 person
Affects Status Importance Assigned to Milestone
mdadm (Ubuntu)
High
Unassigned
Xenial
Undecided
Unassigned
Bionic
Undecided
Unassigned
Cosmic
Undecided
Unassigned
Disco
High
Unassigned

Bug Description

Currently the mdadm package will install a checkraid crontab, that is the same on all installations. This is fine for a single installation, but in bigger (distributed) environments this hurts performance.
E.g. having mysql clusters on softRaid all having their arrays checked at the same moment, causes problems with availability.

An solution would / could be to add an option that will allow the date-time to be randomized. Then this option / question can be used where appropriate, while still retaining the option to have the static configuration as a default.

tags: added: field-medium
Changed in mdadm (Ubuntu):
status: New → Triaged
importance: Undecided → High
tags: added: rls-dd-incoming
summary: - crontab is fixed on specific moment
+ mdadm checkraid cronjob runs at the same time
Steve Langasek (vorlon) on 2019-02-14
tags: removed: rls-dd-incoming
tags: added: id-5c6596886f8e66197bce7d24
Launchpad Janitor (janitor) wrote :

This bug was fixed in the package mdadm - 4.1-2ubuntu3

---------------
mdadm (4.1-2ubuntu3) eoan; urgency=medium

  * Drop /etc/cron.d/mdadm and migrate to systemd mdcheck_start|continue
    timer units.
  * Drop /etc/cron.daily/mdadm and migrate to system mdmonitor-oneshot
    timer unit.
  * mdcheck_start.timer configures the mdcheck on a first sunday of the
    month, with a randomized start delay of up to 24h, and runs for at
    most 6h. mdcheck_continue.timer kicks off daily, with a randomized
    start delay of up to 12h, and continues mdcheck for at most 6h.
  * mdmonitor-oneshot.timer runs daily, with a randomized start delay of
    up to 24h.
  * One can use systemd drop-ins to change .timer units timings, set
    environmental variables to decrease/increase the length of checking,
    or start the checks by hand. Previously used checkarray is still
    available, albeit not used by timer units.
  * Above ensures that previous daily / monthly checks are performed, but
    are randomized, such that performance is not as impacted across a
    cluster of machines. LP: #1815201

 -- Dimitri John Ledkov <email address hidden> Tue, 18 Jun 2019 16:29:44 +0100

Changed in mdadm (Ubuntu):
status: Triaged → Fix Released
To post a comment you must log in.
This report contains Public information  Edit
Everyone can see this information.

Other bug subscribers