Spring 2007.1 Boxed Version Available - Mandriva

This is a discussion on Spring 2007.1 Boxed Version Available - Mandriva ; Bit Twister wrote: > One machine is the server and one is the client. Ok. gw is the server. > On the server machine you export the directories to be mounted/read by > the client. To keep it simple: [root@gw ...

+ Reply to Thread
Page 2 of 2 FirstFirst 1 2
Results 21 to 37 of 37

Thread: Spring 2007.1 Boxed Version Available

  1. Re: Spring 2007.1 Boxed Version Available

    Bit Twister wrote:
    > One machine is the server and one is the client.


    Ok. gw is the server.

    > On the server machine you export the directories to be mounted/read by
    > the client.


    To keep it simple:

    [root@gw etc]# cat exports
    / gw(rw)
    /etc gw(rw)
    /home gw(rw)
    /var gw(rw)
    /mnt/windows gw(rw)

    On jb, /xr /xhome /xvar /xwindows created.

    > I have a script to mount fw:/* directories on wb:/x* mount points.
    >
    >
    > _options="rsize=8192,wsize=8192,timeo=14,intr"
    > _dir=(/ /site /accounts /local)
    > _mpoint=(/xr /xsite /xaccounts /xlocal) /* altered to names above */
    >
    > # services which should be running on both systems
    >
    > service nfslock start
    > service nfs start
    > service netfs start
    > service portmap start


    All that is running on both.
    >
    > i=0
    > while [ $i -lt ${#_mpoint[@]} ] ; do
    > mount -t nfs -o $_options $_node:${_dir[$i]} ${_mpoint[$i]}
    > i=$(( $i + 1 ))
    > done


    This did not work. What is $_node ?
    I do not see that in your script.

    Cheers!

    jim b.

    --
    Unix is not user-unfriendly; it merely
    expects users to be computer-friendly.

  2. Re: Spring 2007.1 Boxed Version Available

    James D. Beard wrote:
    > Bit Twister wrote:
    >> One machine is the server and one is the client.

    >
    > Ok. gw is the server.
    >
    >> On the server machine you export the directories to be mounted/read by
    >> the client.

    >
    > To keep it simple:
    >
    > [root@gw etc]# cat exports
    > / gw(rw)
    > /etc gw(rw)
    > /home gw(rw)
    > /var gw(rw)
    > /mnt/windows gw(rw)
    >
    > On jb, /xr /xhome /xvar /xwindows created.
    >
    >> I have a script to mount fw:/* directories on wb:/x* mount points.
    >>
    >>
    >> _options="rsize=8192,wsize=8192,timeo=14,intr"
    >> _dir=(/ /site /accounts /local)
    >> _mpoint=(/xr /xsite /xaccounts /xlocal) /* altered to names above */
    >>
    >> # services which should be running on both systems
    >>
    >> service nfslock start
    >> service nfs start
    >> service netfs start
    >> service portmap start

    >
    > All that is running on both.
    >>
    >> i=0
    >> while [ $i -lt ${#_mpoint[@]} ] ; do
    >> mount -t nfs -o $_options $_node:${_dir[$i]} ${_mpoint[$i]}
    >> i=$(( $i + 1 ))
    >> done

    >
    > This did not work. What is $_node ?
    > I do not see that in your script.
    >
    > Cheers!
    >
    > jim b.
    >

    Yippee!

    I put gw in the exports file on gw, when it should have be jb.
    _node has to be gw or the ip address. I have not made the
    script run yet, but mounted a directory with
    mount -t nfs gw:/home /xhome

    I tried doing a
    cp -pR /xhome/tree
    and some of the stuff copied over, but some was permission
    denied. I will deal with this later. Any advice appreciated
    but I will not look at it tonight.

    Cheers!

    jim b.

    --
    Unix is not user-unfriendly; it merely
    expects users to be computer-friendly.

  3. Re: Spring 2007.1 Boxed Version Available

    On Wed, 16 May 2007 01:03:19 GMT, James D. Beard wrote:
    >
    > This did not work. What is $_node ?
    > I do not see that in your script.


    Sorry, those instructions were a chop from several scripts.

    I have differnet nodes/pcs, _node would be whichever pc I would be hauling
    files down for backup.

    You would just change $_node to whichever node has the /etc/export
    file.

  4. Re: Spring 2007.1 Boxed Version Available

    On Wed, 16 May 2007 01:40:43 GMT, James D. Beard wrote:
    > Yippee!
    >
    > I put gw in the exports file on gw, when it should have be jb.
    > _node has to be gw or the ip address.


    If you have node name in /etc/hosts then you would not need ip addy.

    > I tried doing a
    > cp -pR /xhome/tree
    > and some of the stuff copied over, but some was permission denied.


    Yep, I execute the copies logged in via a root terminal.

    > I will deal with this later. Any advice appreciated
    > but I will not look at it tonight.


    It also helps when the UID/GIDs match on both systems when you are wanting
    to access files as a regular user.

  5. Re: Spring 2007.1 Boxed Version Available

    Bit Twister wrote:
    >
    > i=0
    > while [ $i -lt ${#_mpoint[@]} ] ; do
    > mount -t nfs -o $_options $_node:${_dir[$i]} ${_mpoint[$i]}
    > i=$(( $i + 1 ))
    > done


    Seems like reinventing the wheel.

    Why not use "noauto" in fstab, or even
    better "autofs".

  6. Re: Spring 2007.1 Boxed Version Available

    On Wed, 16 May 2007 17:01:51 +1000, foo wrote:
    >
    > Seems like reinventing the wheel.
    >
    > Why not use "noauto" in fstab, or even
    > better "autofs".


    I would have to have each pc on the LAN with a different mount point
    in fstab on whichever pc I use for backup and in whatever release I
    happen to be running.

    $ grep label /etc/lilo.conf
    label="/2007_1"
    label="/2007_0"
    label="/ulteo"
    label="/2006oe"
    label="/suse"
    label="/kubuntu"
    label="/fc5"
    label="/ubuntu"


    To add or delete a pc, all I have to do is change the _pc_list
    for _node in $_pc_list ; do
    (mount point loop script here)
    done
    in the backup script. No dinking with fstab(s).

    Guessing with the SATA drives looking like scsi devices, I will have
    to move several of my shared partitions under a common partition
    to free up partition slots. :-(

    As I misunderstand it, we only get 16 partitions on a scsi drive, which
    kinda bites on a 200 gig SATA drive.

    Putting the work in a script keeps node changes to only
    one script file instead of maintaining several /etc/fstabs when I
    add/delete a node/release/distribution.

    Also helps when I clone the setup to someone I setup on linux and
    they do not want to keep my node names.

  7. Re: Spring 2007.1 Boxed Version Available

    On Wed, 16 May 2007 06:20:16 -0400, Bit Twister wrote:

    > As I misunderstand it, we only get 16 partitions on a scsi drive, which
    > kinda bites on a 200 gig SATA drive.


    If all of your systems support it, I'd use LVM, to get around the limit.

    Regards, Dave Hodgins

    --
    Change nomail.afraid.org to ody.ca to reply by email.
    (nomail.afraid.org has been set up specifically for
    use in usenet. Feel free to use it yourself.)

  8. Re: Spring 2007.1 Boxed Version Available

    On Wed, 16 May 2007 17:24:11 -0400, David W. Hodgins wrote:
    > On Wed, 16 May 2007 06:20:16 -0400, Bit Twister wrote:
    >
    >> As I misunderstand it, we only get 16 partitions on a scsi drive, which
    >> kinda bites on a 200 gig SATA drive.

    >
    > If all of your systems support it, I'd use LVM, to get around the limit.


    Yes, and there is the rub.

    Also I thought I saw someone complain that Windows clobbers LVM partitions.

  9. Re: Spring 2007.1 Boxed Version Available

    On Wed, 16 May 2007 18:15:04 -0400, Bit Twister wrote:

    > Also I thought I saw someone complain that Windows clobbers LVM partitions.


    I am the one who posted the warning, however the warning wasn't that m$ clobbers
    LVM partitions, the warning was that it clobbers LVM drives. The difference is important,
    but I guess I wasn't clear enough, when I posted the warning.

    An LVM physical volume may be stored in an LVM partition, or in an LVM drive.
    An LVM partition has an entry in the partition table with the type set to 8e (Linux LVM).
    LVM drives do not have a partition table. The entire drive, including the mbr is used
    as a LVM physical volume. M$ clobbers the mbr, if it doesn't find a partition table.

    It is safe to use LVM partitions on a system that includes an m$ os.
    It is not safe to use LVM drives on a system that includes an m$ os.

    If you look at man pvcreate, the example it shows includes setting up
    an lvm partition, and an lvm drive with "pvcreate /dev/sdc4 /dev/sde"

    Don't use an entire device like /dev/sde, if you will ever boot an m$ os with the device
    connected. In my case, I was experimenting
    with pvcreate /dev/sda, which was a usb flash drive. I didn't think to unplug
    it when I did my monthly boot into xp, to get the latest updates, and xp
    overwrote the first sector, with a blank partition table.

    Even though the physical drive did not contain any actual data, so recovery should
    have been easy, my first attempt ended up wiping out the entire volume group. I'm
    not sure what the proper procedure would have been to safely recover from the lost
    physical drive, and haven't felt like recreating the problem, to figure it out.

    Luckily, the volume group I'd added the lvm physical drive into, was the one I was
    using for my backup/experiment copy of linux, not the one I was regularly using.
    I also had a full backup of the entire system, but as I was about to install 2007.1,
    I didn't bother recovering the lost volume group.

    Some day, when I get bored, I will recreate the problem, and hopefully figure out
    the proper way to recover from it. If/when I do, I'll post the results here.

    Regards, Dave Hodgins

    --
    Change nomail.afraid.org to ody.ca to reply by email.
    (nomail.afraid.org has been set up specifically for
    use in usenet. Feel free to use it yourself.)

  10. Re: Spring 2007.1 Boxed Version Available

    On Wed, 16 May 2007 18:54:10 -0400, David W. Hodgins wrote:
    > On Wed, 16 May 2007 18:15:04 -0400, Bit Twister wrote:
    >
    >> Also I thought I saw someone complain that Windows clobbers LVM partitions.

    >
    > I am the one who posted the warning, however the warning wasn't that
    > m$ clobbers LVM partitions, the warning was that it clobbers LVM
    > drives. The difference is important, but I guess I wasn't clear
    > enough, when I posted the warning.


    You may have been clear enough. I assumed partitions because after my
    2'nd Tue boot, I notice that linux is "fixing partition time in the
    future" on all my partitions, so I assumed Micro$not was dinking with all
    partitions.


  11. Re: Spring 2007.1 Boxed Version Available

    Bit Twister wrote:
    > On Wed, 16 May 2007 01:03:19 GMT, James D. Beard wrote:
    >> This did not work. What is $_node ?
    >> I do not see that in your script.

    >
    > Sorry, those instructions were a chop from several scripts.
    >
    > I have differnet nodes/pcs, _node would be whichever pc I would be hauling
    > files down for backup.
    >
    > You would just change $_node to whichever node has the /etc/export
    > file.


    Yep. That would do it. But for the moment, I am simply
    mounting them the defaults way:
    mount -t nfs gw:/bin /xbin
    with bin replaced by r (for root directory) or the bin name.

    Your earlier mention of making the user id numbers match
    jogged my memory. The install did not give me a choice, and
    I did have a mismatch. I used userdrake to delete the two
    users I had created (without deleting their home directory
    trees), then recreated them setting the id number manually,
    and then as root cd /home ; chown -R user1:user1 user1 ;
    chown -R user2:user2 user2 to get everything properly aligned.

    I had first tried changing the numbers in /etc/passwd, but
    that does not change the shadow password file (which I did not
    tinker with) so I simply shifted to the delete/recreate route.

    Other things: In the router, the original two-line
    Action Name Source Dest Protocol
    Deny Default *,* LAN,* *,*
    Allow Default LAN,* *,* *,*
    allows connectivity among all machines on the local net.
    This was not my problem, but I did not know that.

    I have been deleting modifications made to the shorewall
    scripts to find out just how much is necessary, and results
    have been other than expected. To ping the new machine,
    I have to have an ACCEPT line on it by ip number somewhere,
    whereas this is not necessary for the old machine. Different
    versions of shorewall, so maybe they will have to be configured
    slightly differently. Or maybe something is still slightly
    out of kilter.

    I would like to use the hosts file and create two networks
    on the same eth0 interface, and then accept traffic that is
    loc to loc while leaving the local to Internet network alone.
    I am not sure this is logically possible. For the moment,
    I am using net:192... as a nodename for one of the machines,
    and it works.

    Odd human factors consideration: I have been using my
    Gateway 1.6 GHz Pentium 4 for maybe 6 years now, and speed
    was fine. But now I get up from my new Microtel with two
    dual-core 64-bit AMD 5000+ cpus, SATA drives, etc, sit
    down at the old Gateway, and it is so slow it is painful!

    I had planned to put 2007.1 on the old machine and keep it as a
    backup, but 128MB RAM will not cut it, and I am not sure it is
    worth the $170 or so it would cost for a pair of PC-800
    DDR modules, given the "slow speed" I am now experiencing.

    For the moment, I need to copy over a bunch of files and
    start getting the new machine properly customized for my
    use.

    Many thanks especially to BitTwister but to all who contributed.
    I will be back later, I am sure, but at the moment I know
    enough to stay busy for a while.

    Cheers!

    jim b.

    --
    Unix is not user-unfriendly; it merely
    expects users to be computer-friendly.

  12. Re: Spring 2007.1 Boxed Version Available

    On Wed, 16 May 2007 23:20:55 GMT, James D. Beard wrote:
    >
    > Your earlier mention of making the user id numbers match
    > jogged my memory. The install did not give me a choice, and
    > I did have a mismatch.


    Yep, and starting number has changed from 500 to 501 and now back to
    500.

    On install, I create a junk account. After login I append my other accounts
    which start at 1500. I can also append them on other Distributions
    which may start at 1000 or something else.

    #!/bin/bash
    #************************************************* *****************
    #*
    #* save_1500 - Saves 15?? uid/gid into /etc/*_1500
    #*
    #* saves 15?? uid/gid /etc/passwd,group,gshadow,shadow to /etc/*_1500
    #*
    #* To append/restore results after install
    #* cat /etc/group_1500 >> /etc/group
    #* cat /etc/gshadow_1500 >> /etc/gshadow
    #* cat /etc/passwd_1500 >> /etc/passwd
    #* cat /etc/shadow_1500 >> /etc/shadow
    #*
    #************************************************* *****************
    typeset -i i=o

    _ifs_bkup="${IFS}"

    function ld_ary
    {
    #**********************************************
    #*
    #* ld_ary - load array[] with first name from
    #* each line from /etc/arg1_1500
    #*
    #* End of array[] contains --
    #*
    #**********************************************

    _ary_fn=/etc/${1}_1500

    IFS=":"
    i=0

    while read line ; do
    set -- $line
    array[$i]=$1
    i=$(( $i + 1 ))
    done < $_ary_fn

    array[$i]="--"

    IFS="${_ifs_bkup}"

    } # end ld_ary

    function gen_1500
    {
    #**********************************************
    #*
    #* gen_1500 - generate /etc/arg1_1500.
    #*
    #* reads /etc/arg1 file and any lines starting with
    #* a name found in array[] are written to
    #* /etc/arg1_1500 with the supplied arg2 perm.
    #*
    #**********************************************

    _in_fn=/etc/$1
    _out_fn=/etc/${1}_1500

    /bin/cp /dev/null $_out_fn
    chmod $2 $_out_fn

    while read line ; do
    _parse=$line
    IFS=":"
    set -- $_parse
    IFS="${_ifs_bkup}"
    i=0
    while [ "${array[$i]}" != "--" ] ; do
    if [ "${array[$i]}" = "$1" ] ; then
    echo $line >> $_out_fn
    break
    fi
    i=$(( $i + 1 ))
    done

    done < $_in_fn

    IFS="${_ifs_bkup}"

    } # end gen_1500

    #***********************************
    #*
    #* Main code start here
    #*
    #***********************************

    grep ':15' /etc/group | grep -v ':15:' | sort -n -t ':' --key=3 > /etc/group_1500
    grep ':15' /etc/passwd | grep -v ':15:' | sort -n -t ':' --key=3 > /etc/passwd_1500

    ld_ary group
    gen_1500 gshadow 600

    ld_ary passwd
    gen_1500 shadow 600

    #**************** end save_1500 **********************************

    > I used userdrake to delete the two
    > users I had created (without deleting their home directory
    > trees), then recreated them setting the id number manually,
    > and then as root cd /home ; chown -R user1:user1 user1 ;
    > chown -R user2:user2 user2 to get everything properly aligned.


    Small gotcha when you do not fix other locations containg user files.
    Example:
    ls -al /var/spool/mail/



    > I had first tried changing the numbers in /etc/passwd, but
    > that does not change the shadow password file (which I did not
    > tinker with) so I simply shifted to the delete/recreate route.


    Hehe, it's just a text file.



    >
    > I would like to use the hosts file and create two networks
    > on the same eth0 interface, and then accept traffic that is
    > loc to loc while leaving the local to Internet network alone.
    > I am not sure this is logically possible.


    Me neither. I know you put a zone on a nic. So, on eth1 I have net for
    the internet, and eht0 is loc for the local lan.
    On a single nic lan pc it would be net: on the firewall box it would
    be loc:

    > For the moment,
    > I am using net:192... as a nodename for one of the machines,
    > and it works.


    See above.

    >
    > Odd human factors consideration: I have been using my
    > Gateway 1.6 GHz Pentium 4 for maybe 6 years now, and speed
    > was fine. But now I get up from my new Microtel with two
    > dual-core 64-bit AMD 5000+ cpus, SATA drives, etc, sit
    > down at the old Gateway, and it is so slow it is painful!


    Hmmm, my firewill
    $ grep MHz /proc/cpuinfo
    cpu MHz : 451.104

    is feeding my browsing lan pc

    $ grep MHz /proc/cpuinfo
    cpu MHz : 2400.000

    and according to http://www.broadbandreports.com/stest I get
    download / upload : 6529 / 482 (Kbps) according to


  13. Re: Spring 2007.1 Boxed Version Available

    On Wed, 16 May 2007 19:18:08 -0400, Bit Twister wrote:

    > You may have been clear enough. I assumed partitions because after my
    > 2'nd Tue boot, I notice that linux is "fixing partition time in the
    > future" on all my partitions, so I assumed Micro$not was dinking with all
    > partitions.


    Could you clarify this a bit? I have no idea what you mean by
    fixing the partition time.

    Regards, Dave Hodgins

    --
    Change nomail.afraid.org to ody.ca to reply by email.
    (nomail.afraid.org has been set up specifically for
    use in usenet. Feel free to use it yourself.)

  14. Re: Spring 2007.1 Boxed Version Available

    On Thu, 17 May 2007 04:30:18 -0400, David W. Hodgins wrote:
    >
    > Could you clarify this a bit? I have no idea what you mean by
    > fixing the partition time.


    Hmmm, after a bit of checking, it seems the file system driver
    (Ext2IFS), which allows doze apps to see linux ext3/2 directories, was
    the culprit. http://uranus.it.swin.edu.au/~jn/linux/ext2ifs.htm

    I had installed the file system driver so I would not have to
    create a fat partition to get linux files to doze.

    I do not know the actual details, but we know doze does NOT play well
    when you have a dual boot system with the hardware clock set GMT.

    In my case hardware clock is set GMT and as far as doze is concerned,
    local time is 6 hours in the future. It appears ext2ifs was setting
    something in my linux partitions with the +6 hr time value.

    Because of that, I was always seeing the time repair messages after booting
    doze and I had assumed it was winders dinking with the linux partitions.


  15. Re: Spring 2007.1 Boxed Version Available

    On Thu, 17 May 2007 09:29:58 -0400, Bit Twister wrote:

    > Hmmm, after a bit of checking, it seems the file system driver
    > (Ext2IFS), which allows doze apps to see linux ext3/2 directories, was
    > the culprit. http://uranus.it.swin.edu.au/~jn/linux/ext2ifs.htm


    A read only driver is updating the timestamps on the file system? Ouch!

    > I do not know the actual details, but we know doze does NOT play well
    > when you have a dual boot system with the hardware clock set GMT.
    > In my case hardware clock is set GMT and as far as doze is concerned,
    > local time is 6 hours in the future. It appears ext2ifs was setting
    > something in my linux partitions with the +6 hr time value.
    > Because of that, I was always seeing the time repair messages after booting
    > doze and I had assumed it was winders dinking with the linux partitions.


    Because I still have 98, and XP on the same computer, I keep the hardware
    time in localtime. Saves a lot of headaches. I have
    $ cat /etc/sysconfig/clock
    UTC=false
    ARC=false
    ZONE=Canada/Eastern

    I'll take a look at the source for ext2ifs, and see if I can figure out
    what's happening.

    Regards, Dave Hodgins

    --
    Change nomail.afraid.org to ody.ca to reply by email.
    (nomail.afraid.org has been set up specifically for
    use in usenet. Feel free to use it yourself.)

  16. Re: Spring 2007.1 Boxed Version Available

    On Thu, 17 May 2007 13:47:52 -0400, David W. Hodgins wrote:
    >
    > A read only driver is updating the timestamps on the file system? Ouch!


    Yes I did not think that was the problem.


    > I'll take a look at the source for ext2ifs, and see if I can figure out
    > what's happening.


    Not unless you want to. Since I have removed it from both systems,
    the message has not shown up. Seems with it installed, the AV sofware
    wants to check all the linux files. That causes it to take about 3
    hours for a scan.

  17. Re: Spring 2007.1 Boxed Version Available

    On Thu, 17 May 2007 14:32:58 -0400, Bit Twister wrote:

    > On Thu, 17 May 2007 13:47:52 -0400, David W. Hodgins wrote:
    >> I'll take a look at the source for ext2ifs, and see if I can figure out
    >> what's happening.

    >
    > Not unless you want to. Since I have removed it from both systems,
    > the message has not shown up. Seems with it installed, the AV sofware
    > wants to check all the linux files. That causes it to take about 3
    > hours for a scan.


    Nothing obvious in the source. The problem must be in a lower level
    m$ kernel i/o module.

    I have similar problem, using captive-fuse to mount an ntfs filesytem,
    as r/w, under linux. When msec-find, or mlocate scans the drive,
    the captive program (running the m$ ntfs modules) seems to have a
    memory leak, and will eventually cause my system to run out of swap
    space. Because of this, I use noauto on the ntfs filesystem, and
    make sure I don't leave it mounted when cron.daily, etc, run.

    Regards, Dave Hodgins

    --
    Change nomail.afraid.org to ody.ca to reply by email.
    (nomail.afraid.org has been set up specifically for
    use in usenet. Feel free to use it yourself.)

+ Reply to Thread
Page 2 of 2 FirstFirst 1 2