LSM swap space science 2

From: Markus Waldorf (waldorfm_at_rferl.org)
Date: 07/02/03

  • Next message: Dr J Pelan: "SUMMARY: Tru64 v5.1B: NFS Server Crash & kernel panic" 
    Date: Wed, 02 Jul 2003 11:39:53 +0200
To: Unix Managers <tru64-unix-managers@ornl.gov>

    sorry, one more add-on. According to the LSM docuementation:

    4.2.6    Creating an LSM Volume for Secondary Swap Space

    It uses the command:

    #volassist -U gen make vol_swap2 128m nmirror=2 layout=nolog

    If I understand correctly, the above creates a "generic" fstype volume. Now
    I wonder why it is not swap. Can only the primary swap volume be type "swap"
    and any futher ones type "gen"?

    Hello,

    I have to admit that LSM has not done a very good job so far to make me more
    comfortable with what I'm doing. I've been playing around with LSM for
    weeks, and I better not mention how many times I decided to start over and
    over again. I've been going through the documentation, best practices and
    mailing lists items, but I'm still searching for some clear answers. I
    believe having the filesystems configured properly is not a trivial issue,
    in particular LSM. Readme later.

    I'm searching for answers how to properly setup LSM swap space. So far I've
    come to the following conclusions. I go somewhere, but perhaps someone could
    let me know if I'm at the right place.

    1) swap volumes must belong to the rootdg disk group and must be nopriv
    disks - at least according to "volencap". The volencap command can be used
    to "encapsulate" a disk (whatever that means), or in other words, it makes
    nopriv disks and creates LSM volumes. Where disk can be a whole disk or just
    a partition.

    For example, I created partitions dsk4b, dsk5b, all type unused, and want to
    use or prepare them for LSM swap mirror:

    # volencap dsk4b dsk5b
    # volreconfig

    The above will automatically create plexes and nopriv LSM volumes in the
    rootdg group. But what about swap space parameters? I cannot use the LSM GUI
    to change the volumes to these parameters, even if the volumes are stopped -
    it results just in errors about bad parameter combinations.

    2) I was reading that swap space should be set no "no recover" and "no log".
    So I did the following in order to manually remove volumes from LSM and
    create them using command-line tools:

    >From the LSM gui advanced menu:

    - working on dsk4b
    - select volume, volume stop
    - select plex (big box) and move it off the volume
    - select volume (now empty) and use remove volume from the menu

    I don't like the default names so I right-click on the plex and the subdisk
    in order to make it more meaningful, e.g. dsk4b: Plex swapvol05-01, subdisk:
    swap05-01.

    Than I issued the following commands to create a new volume named swapvol03
    with the desired options:

    # volmake -U swap vol swapvol03 plex=swapvol05-01 log_type=NONE
    # volume set start_opts=norecov swapvol03

    I want to mirror dsk4b and dsk5b, so back in LSM gui:

    >From the LSM gui advanced menu:

    - working on dsk5b
    - select volume, volume stop
    - select plex (big box) and move it off the volume
    - select volume (now empty) and use remove volume from the menu
    - select subvolume within plex and move it off with the mouse
    - select plex and remove it
    - select subvolume on disk and remove it

    - select swapvol03 and middle-click dsk5b
    - from the basic LSM gui menu select add mirror
    - wait for it to complete
    - rename the subdisk and plex by right-clicking them

    Well, it seems to work, but what It adds another subdisk to each of the the
    swapvol03 plexes. Is that ok? I did not see this when using ie. "volrootmir"

    Here is a volprint:

    # volprint -htl swapvol03
    Disk group: rootdg

    Volume: swapvol03
    info: len=8912896
    type: usetype=swap
    state: state=ACTIVE kernel=ENABLED
    assoc: plexes=swapvol05-01,swapvol03-01
    policies: read=ROUND exceptions=GEN_DET_SPARSE
    flags: closed writeback
    logging: type=NONE
    apprecov: seqno=0
    device: minor=1 bdev=40/1 cdev=40/1 path=/dev/vol/rootdg/swapvol03
    perms: user=root group=system mode=0600
    options: norecov

    Plex: swapvol05-01
    info: len=8912896
    type: layout=CONCAT
    state: state=ACTIVE kernel=ENABLED io=read-write
    assoc: vol=swapvol03 sd=swap05-01p,swap05-01
    flags: complete

    Subdisk: swap05-01p
    info: disk=dsk4b offset=0 len=16
    assoc: vol=swapvol03 plex=swapvol05-01 (offset=0)
    flags: enabled (nullwr)
    device: device=dsk4b path=/dev/disk/dsk4b diskdev=81/257

    Subdisk: swap05-01
    info: disk=dsk4b offset=16 len=8912880
    assoc: vol=swapvol03 plex=swapvol05-01 (offset=16)
    flags: enabled
    device: device=dsk4b path=/dev/disk/dsk4b diskdev=81/257

    Plex: swapvol03-01
    info: len=8912896
    type: layout=CONCAT
    state: state=ACTIVE kernel=ENABLED io=read-write
    assoc: vol=swapvol03 sd=dsk5b-01p,dsk5b-01
    flags: complete

    Subdisk: dsk5b-01p
    info: disk=dsk5b offset=0 len=16
    assoc: vol=swapvol03 plex=swapvol03-01 (offset=0)
    flags: enabled (nullwr)
    device: device=dsk5b path=/dev/disk/dsk5b diskdev=81/321

    Subdisk: dsk5b-01
    info: disk=dsk5b offset=16 len=8912880
    assoc: vol=swapvol03 plex=swapvol03-01 (offset=16)
    flags: enabled
    device: device=dsk5b path=/dev/disk/dsk5b diskdev=81/321

    # volprint -hlt swapvol03
    Disk group: rootdg

    Volume: swapvol03
    info: len=8912896
    type: usetype=swap
    state: state=ACTIVE kernel=ENABLED
    assoc: plexes=swapvol05-01,swapvol06-01
    policies: read=ROUND exceptions=GEN_DET_SPARSE
    flags: closed writeback
    logging: type=NONE
    apprecov: seqno=0
    device: minor=1 bdev=40/1 cdev=40/1 path=/dev/vol/rootdg/swapvol03
    perms: user=root group=system mode=0600
    options: norecov

    Plex: swapvol05-01
    info: len=8912896
    type: layout=CONCAT
    state: state=ACTIVE kernel=ENABLED io=read-write
    assoc: vol=swapvol03 sd=swap05-01p,swap05-01
    flags: complete

    Subdisk: swap05-01p
    info: disk=dsk4b offset=0 len=16
    assoc: vol=swapvol03 plex=swapvol05-01 (offset=0)
    flags: enabled (nullwr)
    device: device=dsk4b path=/dev/disk/dsk4b diskdev=81/257

    Subdisk: swap05-01
    info: disk=dsk4b offset=16 len=8912880
    assoc: vol=swapvol03 plex=swapvol05-01 (offset=16)
    flags: enabled
    device: device=dsk4b path=/dev/disk/dsk4b diskdev=81/257

    Plex: swapvol06-01
    info: len=8912896
    type: layout=CONCAT
    state: state=ACTIVE kernel=ENABLED io=read-write
    assoc: vol=swapvol03 sd=swap06-01p,swap06-01
    flags: complete

    Subdisk: swap06-01p
    info: disk=dsk5b offset=0 len=16
    assoc: vol=swapvol03 plex=swapvol06-01 (offset=0)
    flags: enabled (nullwr)
    device: device=dsk5b path=/dev/disk/dsk5b diskdev=81/321

    Subdisk: swap06-01
    info: disk=dsk5b offset=16 len=8912880
    assoc: vol=swapvol03 plex=swapvol06-01 (offset=16)
    flags: enabled
    device: device=dsk5b path=/dev/disk/dsk5b diskdev=81/321

    #

  • Next message: Dr J Pelan: "SUMMARY: Tru64 v5.1B: NFS Server Crash & kernel panic"

    Relevant Pages

    • LSM swap space science
      ... I have to admit that LSM has not done a very good job so far to make me more ... to "encapsulate" a disk, or in other words, it makes ... assoc: plexes=swapvol05-01,swapvol03-01 ... flags: closed writeback ...
      (Tru64-UNIX-Managers)
    • Swap mirroring with LSM
      ... I'm wondering if there would be a performance hit from mirroring my swap volume in LSM. ... Intuitively, mirrored swap doesn't sound like it would be very performance-friendly, however this depends I think on how the LSM driver handles I/O. ... Unlimited Netscape Internet Service. ...
      (Tru64-UNIX-Managers)