SUMMARY:cluster_root, cluster_usr & cluster_var storage

From: ^IT_MPG_UNIX_ADMIN <^IT__MPG__UNIX__ADMIN_at_ccmail.allegiance.net>
Date: Tue, 29 May 2001 12:31:10 -0500

     Many thanks to the following people -
     
     Jefferey Hummel Jeffrey_Hummel_at_albemarle.com
     Kathee kathee_at_mindiq.com
     Scott Fafrak scott.fafrak_at_bruden.com
     Brad Nichols brad.nichols_at_compaq.com
     Chia-Jen Chyan chyan_at_decatl.alf.dec.com
     
     The answer is "YES". I include a brief summary below -
     
     Yes, cluster_root, cluster_usr and cluster_var have no special advfs
     restrictions - unlike a root_domain in a standalone system which must
     be the 'a' partition and cannot have multiple volumes (no addvol or
     rmvol). So yes you can add a volume then remove the original volume
     the same as any other advfs domain - but on that note - expect that
     the rmvol will fail if their is not enough remaining space for the
     currently defined files. I wouldn't do it under load if I didn't have
     to (wait for off hours.)
     
     On the other hand, the more complex you make the data definition for
       the file systems the more error prone and/or hard to restore
       they become - i.e. having 20 tiny volumes as your
       cluster_root~! . So the suggestion is keep it simple and wait
     for performance to be a measurable problem before going to far.
     
     Tks & Rgds,
     
     Alay Shah
     Allegiance Healthcare Corporation
     1400 Waukegan Rd.
     McGaw Park, IL 60085
     Ph - 847-578-2584
     Fax - 847-578-5586
     Email - unixadmin_at_allegiance.net
     shahal_at_allegiance.net
     
     
     
             
     
     
     ----------------------ORIGINAL
     QUESTION------------------------------------
     
     When I created our V5.1 cluster, I put all three domains on different
     slices of the same mirror set. Here's the disklabel from that disk
     
     a: 1048576 0 AdvFS # b:
     1039373 1048576 unused 0 0 # c:
     35556389 0 unused 0 0 # d:
     4096 2087949 unused 0 0 # e: 0
            0 unused 0 0 # f: 0
     0 unused 0 0 # g: 16732172 2092045
     AdvFS # h: 16732172 18824217 AdvFS
                               #
     
     where a is cluster_root, g is cluster_usr and h is cluster_var. But
     now I am concerned about performance because I will have 6 cluster
     members in this cluster eventually trying to hit /var/tmp on the same
     cluster_var. Also, lots of our applications put data out in /usr/opt
     which is in /usr. Also my /usr filesystem is already at 55% full. What
     I should have really done is created raidsets(RAID 3/5) and put /,
     /usr and /var on their own raidsets. In any case, I know that we can
     bump up the storage by adding volumes using addvol. But can I remove
     the old disk using rmvol? What I would like to do is add the new raid
     sets and remove the old disk partitions that I initially defined. Any
     comments/suggestions/information will be greatly appreciated.
     
     Tks & Rgds,
     
     Alay Shah
     Allegiance Healthcare Corporation
     1400 Waukegan Rd.
     McGaw Park, IL 60085
     Ph - 847-578-2584
     Fax - 847-578-5586
     Email - unixadmin_at_allegiance.net
     shahal_at_allegiance.net
     
Received on Tue May 29 2001 - 17:38:52 NZST

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:42 NZDT