Unix Admin Notes

This is a miscellaneous collection of things I want to remember when administering my machines. These are notes for me managing my Debian systems, so there's little effort made to explain things (more than I need to know). Maybe someone else will find something useful here too, but you should nothing here as completely accurate.

• What's in an ISO image?

    sudo mount -o loop -t iso9660 /tmp/floppy.iso /mnt/iso
    mkdir /tmp/jjj
    sudo mount -o loop /mnt/iso/boot/boot.img /tmp/jjj
  
    Now you can see a directory structure of data in /tmp/jjj
  

  
  
• What's in an initrd file? See http://www.ibm.com/developerworks/linux/library/l-initrd.html

    sudo mkdir /dev/shm/initrd
    sudo mount -o loop -t cramfs /boot/initrd.img-2.6.14-mos /dev/shm/initrd
    
    Note you can get a shell just before the pivot is done by changing
    DELAY in /etc/mkinitrd/mkinitrd.conf
  

  
  
• Set up locales fix errors like this:

     perl: warning: Setting locale failed.
     perl: warning: Please check that your locale settings:
        LANGUAGE = "en_US",
        LC_ALL = (unset),
        LANG = "en_US"
     are supported and installed on your system.
     perl: warning: Falling back to the standard locale ("C").
  

  You only need install the package locales. In particular people seem to think localeconf is a bad thing.

   sudo vi /etc/environment    # Should just have LANG="en_US.UTF-8"
  
   sudo vi /etc/locale.gen     # If exists, it should look like:
                               #  en_US.UTF-8 UTF-8
                               #  en_US ISO-8859-1
  
   sudo dpkg-reconfigure locales
   Enable just the two en_US languages above. Default should be 'en_US'
  

  
  
• Snatch X11 screen image from a remote machine.

  Sometimes I'll need to run a LiveCD to do diagnostics or some other maintenance which is long running. Recent versions of the Dell Remote Access Controller (RAC) have their JAVA console broken for 95% of the browsers in the universe. This hack allows me to snatch the X11 screen image so I can tell when the maintenance is complete. Note, this only applies if the remote machine has an X11 session open. It's also proof of why root should normally never run X11 sessions.

    ssh -X remotemachine        # SSH with X support to remote machine
    ls -la /root/.Xauthority    # Xauthority for owner of console session
    export DISPLAY=:0.0         # DISPLAY value for local console
    xauth merge /root/.Xauthority   # Allow access to new DISPLAY
    xwd -root -out /tmp/console.xwd # Grab console of local machine
    scp -p /tmp/console.xwd user@host:/tmp    # Screen shot to MY machine
    gimp /tmp/console.xwd       # View remote console on my machine
  

  
  
• Make package for stable from another level. See http://www.debian-administration.org/articles/20.

    Visit http://www.debian.org/distrib/packages
    Find the package of interest in some other distribution
  
    Move to the package description. In the blue box on the RH side find 'Download Source Package'
      click on and save the *.orig.tar.gz file
      click on and save the *diff.gz file
  
    Move to a place where the source is to be extracted and built
      Note this may be on a machine with the correct libs installed
      cd ~/src
      tar xzvf /tmp/packagename.orig.tar.gz
      cd packagename
      gunzip -c /tmp/packagename*diff.gz | patch -Np1
  
    Make the package  (requires packages devscripts fakeroot)
      debuild -i -us -uc -b
  

  
  
• Set up Debian mirror

    Create mirror and local package place
      cd /home/placeformirror
      mkdir debian
    Find anonftpsync.sh from web. Edit it and set up the TO=
    to /home/placeformirror/debian and RSYNC_HOST=ftp.egr.msu.edu
    and MAILTO="proper@mailtoaddress"
  
    sudo nohup anonftpsync.sh         # No mesages, takes a long time 1st time
  
    Set up a web server like lighttpd. Configure /etc/lighttpd/lighttpd.conf
  
    Start server with sudo /etc/init.d/lighttpd restart
  

  
  
• Create a partition for devices > 2TB See http://www.gnu.org/software/parted/. Careful here, parted makes changes immediately to disk.

    parted
      use /dev/sdc
      unit MB
      print                     # Note size of device
        Disk geometry for /dev/sdc: 5624000MB
        Disk label type: none   # <=== should be msdos (<2TB) or gpt (>2TB)
  
      mklabel msdos  (or gpt)   # Probably latter
  
      # Note that "parted" uses MB for the partition boundaries
      mkpart primary ext3 0 5624000    # Will not use ext3 filesystem  Extra 000
      print                     # Should look like you expect 
      quit
  
  
    After the partition is created, do mkfs.xfs etc as usual
  

  
  
• Set up NFS See http://nfs.sourceforge.net/nfs-howto/.

    Install the packages
      sudo apt-get install nfs-kernel-server
  
    sudo vi /etc/hosts.allow      add this line
      portmap: 192.168.0.0/255.255.0.0
  
    sudo vi /etc/exports
      #   Allow cluster clients to get at NFSROOT
      /home0 192.168.0.0/16(async,ro,no_root_squash,no_subtree_check)
      /home  192.168.0.0/16(async,ro,no_root_squash,no_subtree_check)
  
    Restart server if need be:
      sudo /etc/init.d/nfs-kernel-server restart
  
    To just read the file by hand, send daemon -HUP kill signal
      sudo /etc/init.d/nfs-kernel-server reload
  
    Useful commands:
      exportfs -f          # Flush all data
      exportfs -u          # Unexport all directories
      exportfs -r          # Reexport all directories
      rpcinfo -p           # See if NFS is up
  
    Mount NFS drive on the remote machine
      mount 192.168.1.1:/home  /remotemachine/home
      vi /etc/fstab
        remotemachine:/home  /remotemachine/home  nfs   ro,soft,intr,proto=udp  0    0
  

  
  
• Software RAID setup See http://tldp.org/HOWTO/Software-RAID-HOWTO-5.html.

   Install software packages
     sudo apt-get install mdadm
  
    Create the device.  In theory mdadm --auto can do this, but... (still needed?)
      sudo mknod /dev/md0 b 9 0
  
    Find and allocate the partitions using 'fdisk /dev/sdX'
    Create the RAID device
      sudo mdadm --create --verbose /dev/md0 --auto yes  --level 1
         --raid-devices 2 /dev/sda4 /dev/sdb4
      This can start a initialization of the device in the background
      that can take a very very long time. While this is going on
      the system is highly unusable. Patience grasshopper.
  
    Create filesystem on disk
      sudo mkfs.xfs -f /dev/md0
  
    Create file so device is known at boot to /etc/init.d/mdadm* scripts
      sudo vi /etc/mdadm/mdadm.conf
        DEVICE  /dev/sdX1 /dev/sdX2 
        ARRAY   /dev/md0 devices=/dev/sdX1,/dev/sdX2
  
    Queries to see the device:
      sudo mdadm --detail --scan
      sudo cat /proc/mdstat 
  
    Add the device to /etc/fstab to be automatically mounted at boot
  

  
  
• Set up ATA over Ethernet (AOE) See http://www.howtoforge.com/ata_over_ethernet_debian_etch.

    Be sure kernels on server (where physical disk is to be found)
    and client (where AOE disk will be mounted) have CONFIG_ATA_OVER_ETH set.
      grep ATA_OVER /boot/config-`uname -r`
  
    Load the aoe module: sudo modprobe aoe
      sudo tail /var/log/syslog      # See 'kernel: aoe: AoE v32 initialised'
  
    SERVER
      Install software:  sudo apt-get install vblade
  
      Start daemon to manage the device. Each AoE device is identified by
      a couple Major/Minor, with major between 0-65535 and minor between 0-255.
      AoE is based just over Ethernet on the OSI models so we need to
      indicate which ethernet card we'll use. We export /dev/sda1 with
      a major value of 0 and minor if 1 on the eth0 interface.
  
      vbladed 0 1 eth0 /dev/sda1
  
    CLIENT 
      As above check the kernel has AOE enabled. Get the module loaded.
  
      Install software:  sudo apt-get install aoetools
  
      Find an AOE devices:
        aoe-discover
        aoe-stat
        ls -al /dev/etherd/       # See details on device that was found/created
  
      Treat the new device (/dev/etherd/e0.1) just like any other device
      mkfs.xfs /dev/etherd/e0.1
      mkdir /aoe
      mount -t xfs /dev/etherd/e0.1 /aoe
  

  
  
• Set up LVM combine many partitions together as one 'virtual disk' See http://tldp.org/HOWTO/LVM-HOWTO/ and http://www.linuxquestions.org/questions/linux-newbie-8/unknown-device-with-lvm-292280/

    Install software
      sudo apt-get install lvm2 dmsetup xfsprogs
  
    Partition disks, fdisk each with one partition (e.g. /dev/sdb1)
  
    Create physical volumes
      sudo pvcreate /dev/sdb1
      sudo pvcreate /dev/sdc1
      sudo pvcreate /dev/sdd1
      sudo pvscan
  
    Create a volume group and make it available
      sudo vgcreate fhome /dev/sdb1 /dev/sdc1 /dev/sdd1
      sudo vgchange -a y fhome
      sudo vgdisplay
      sudo lvcreate --size 3493G --name fanhome fhome
      sudo lvdisplay
      mount /dev/bak/backups /BAK/backups
  
    Rename a logical volume group
      sudo vgdisplay -v shome
      sudo vgchange -a n /dev/shome
      sudo vgrename /dev/shome /dev/ext
      sudo vgchange -a y /dev/ext
      sudo vgdisplay -v
      sudo vgdisplay -v ext
  
    Rename a logical volume
      sudo lvdisplay  /dev/ext/snowhome
      sudo lvrename /dev/ext/snowhome /dev/ext/nfsdata
      sudo lvdisplay  /dev/ext/nfsdata
  
    Make a file system on the disks
      sudo mkfs.xfs -f /dev/fhome/fanhome
      sudo mount /dev/fhome/fanhome /home
      sudo mount -t xfs /dev/ext/nfsdata /mnt
  
    And again
      sudo vgcreate shome /dev/sde1
      sudo vgchange -a y shome
      sudo lvcreate --size 698G --name snowhome shome
      sudo mkfs.xfs -f /dev/shome/snowhome
      sudo mount /dev/shome/snowhome /snowhome
  
    Add a new drive to a VG
      sudo vgdisplay -v     # See what disks are used
      # create partition, format as usual
      umount /BAK/backups
      sudo pvcreate /dev/sdf1   # Prepare new partition
      sudo vgextend bak /dev/sdf1
      sudo lvresize --size 5.999T /dev/shome/home    # Increase size 
      sudo mount /dev/bak/backups /BAK/backups
      sudo xfs_growfs /BAK/backups
  

  
  
• Set up iSCSI support See http://www.open-iscsi.org/docs/README.

    Install software. Requires iscsi-tcp should be compiled into kernel.
      sudo apt-get install open-iscsi parted
  
    Find the initiator name for this node:
      tail -1 /etc/initiatorname.iscsi 
  
    Make sure that this initiator is defined to the iSCSI device
      firefox --no-remote http://raid3
      set Host LUN -> Create iSCSI Initiator
  
    Figure out what iSCSI initiators are around and their LUNs
      sudo iscsiadm -m discovery -t sendtargets -p 192.88.99.240:3260
  
    Creates files in /etc/iscsi/nodes AND changes the node.startup value.
    Do not do this unless you correct the startup values.
  
    sudo vi /etc/iscsi/iscsid.conf     #  Make sure 'node.startup = automatic'
  
  * Set up iscsi devices to be started automatically
    cd /etc/iscsi/nodes
    sudo vi */*/default
      change 'node.startup = manual' to 'node.startup = automatic'
  
    The following did not work for me:
      To automate login to a node, use the following with the record ID (record ID
      is the targetname and portal) of the node discovered in the discovery above:
    	iscsiadm -m node -T targetname -p ip:port --op update -n node.conn[0].startup -v automatic
  
    Perhaps restart and check what happened    
      sudo /etc/init.d/open-iscsi restart
      sudo tail -20 /var/log/syslog
  
    Find out what is going on
      sudo iscsiadm -m session
      tcp: [1] 192.168.1.240:3260,1 iqn.2002-10.com.infortrend:raid.sn7350865.00
  
    See statistics about activity
      sudo iscsiadm -m session -s -r ID
  
    Make file system as usual and mount
      mkfs.xfs -f /dev/sdc1