RHEL4 and Disk Druid inconsistencies

From: Sean Hawthorne (shawthorne_at_roam-it.com)
Date: 10/27/05

  • Next message: Michael Velez: "RE: senamil with two domains" 
    Date: Thu, 27 Oct 2005 11:15:12 -0600
To: General Red Hat Linux discussion list <redhat-list@redhat.com>

    Hello All,

    I've been trying my darnedest to build a pair of clustered servers this
    week. I have been though the RHEL4 install on all four machines. All of
    which are the exact same hardware. One thing that is driving me nuts
    though is Disk Druid. I have been very careful about building all my
    partitions exactly the same (same order and sizes, etc....), yet I keep
    getting different results. i.e.

    I'll slice both hard drives like so:

    sda1 200MB software raid
    sda2 2000MB software raid
    sda3 6000MB software raid
    sda4 extended partition
    sda5 232800MB software raid

    All the cluster sizes match and everything. All seems good. Then I start
    making the RAID sets. Some times I'll end up with this:

    md0 200MB /boot
    md1 6000MB /
    md2 2000MB swap
    md3 232800MB /data

    other times I'll end up with this (or something else completely).

    md0 2000MB swap
    md1 6000MB /
    md2 232800MB /data
    md3 200MB /boot

    There doesn't seem to be any consistency in how disk druid orders the md
    devices when they are created. While not 100% necessary to have the
    clusters all setup exactly the same, it certainly does make
    administering them easier if I don't have to deal with nuances like this.

    So, my questions:

    1) Can anyone tell me how to repeatably get the same results from disk
    druid each time, or why it does this?
    2) Is there any way to reorder the md devices once the system is
    installed? What do I need to change/look for? 

    -- 
Sean Hawthorne
Roam IT Technical Professional
Email: shawthorne@roam-it.com
-- 
redhat-list mailing list
unsubscribe mailto:redhat-list-request@redhat.com?subject=unsubscribe
https://www.redhat.com/mailman/listinfo/redhat-list
  • Next message: Michael Velez: "RE: senamil with two domains"