AIR Wiki : ChangeLogRembrandt2005

HomePage :: Categories :: PageIndex :: RecentChanges :: RecentlyCommented :: Login/Register

Changelog Rembrandt Cluster 2005


Modified Files in /etc after System Upgrade

2005-12-01 by Hans Blom
Why: After system upgrade some files were replaced or distribution alternatives were proposed
Where: On nodes "rembrandt0-2,6-8"
How:
The following modification were done:

Moved Home Directories from vangogh0 to rembrandt0

2005-11-29 by Hans Blom
Why: The VanGogh cluster will be upgraded
Where: On all Rembrandt nodes
How:

Clean Up The /etc/network/interfaces Files

2005-11-21 by Hans Blom
Why: These files had become a bit messy by successive configuration changes
Where: On all Rembrandt nodes
How:

Create ARP Addess Mapping Entries for TBN

2005-11-03 by Hans Blom
Why: Some ARP entries should be set for the Token Based Networking tests
Where: On nodes "rembrandt6-8"
How:

Building and Installing Kernel 2.6.14

2005-11-03 by Hans Blom
Why: Contains already driver for Chelsio N210
Where: On nodes "rembrandt0-2,5-8"
How:
cd /usr/src
tar xfvj ~/tar/linux-2.6.14.tar.bz2
rm linux
ln -si linux-2.6.14 linux
cd /usr/src/linux-2.6.14
cp -pi /boot/config-2.6.12.2 .config   # Use existing 2.6.12.2 config.
make menuconfig
# Enable the Chelsio driver as module:
#    Device Drivers  --->
#         Network device support  --->
#             Ethernet (10000 Mbit)  --->
#                  <M> Chelsio 10Gb Ethernet support
cp -pi .config config.save
make bzImage 2>&1 | tee make_bzImage.out
make modules 2>&1 | tee make_modules.out
make modules_install 2>&1 | tee make_modules_install.out
cp -i arch/x86_64/boot/bzImage /boot/vmlinuz-2.6.14
cp -i .config /boot/config-2.6.14
cp -i System.map /boot/System.map-2.6.14
mkinitrd -o initrd.img-2.6.14 2.6.14 2>&1 | tee mkinitrd.out
cp -i initrd.img-2.6.14 /boot
gvim /boot/grub/menu.lst

Use Full Hostname for NFS Mount Tier-1 Hosts

2005-10-06 by Hans Blom
Why: Otherwise no "rembrandt0" NFS mounts possible.
Where: On nodes "rembrandt3,4".
How:
The base hostnames are now used for the Tier-1 test network. Therefore, the home directory server "rembrandt0" should now be fully specified. To do this in /usr/share/am-utils/auto_home_rembrandt0, the phrase rhost:=rembrandt0 is changed into rhost:=rembrandt0.uva.netherlight.nl. After that the auto mounter is restarted by running /etc/init.d/am-utils restart. Note that for home directory server "vangogh0" there are no problems because its base hostname is still used for the management subnet.

Added ramdisk on Rembrandt7 and 8

2005-09-26 by FreekDijkstra
Why: Needed for iGrid demo.
Where: On nodes "rembrandt7,8".
How:
mkdir /space/ramdisk
mount -t tmpfs -o size=1000M,mode=0777 tmpf /space/ramdisk

Note that this is not preserved after reboot, since I did not modify /etc/fstab

Set VLAN Tags

2005-09-19 by Hans Blom
Why: To be able to use at switches also tagged ports.
Where: On nodes "rembrandt1,2".
How:

Load IPv6 Module at Boot to Set Correctly IPv6 Static Routes

2005-09-19 by Hans Blom
Why: The static IPv6 addresses were not configured at system boot.
Where: On nodes "rembrandt1,2".
How:

Install Driver Module for Chelsio N210 NIC in 2.6.12.2 Kernel

2005-08-23 by Hans Blom
Why: To be able to use the Chelsio N210 NICs.
Where: On nodes "rembrandt1,2".
How:

Install 2.6.12.2 Kernel at Hosts where SATA is /dev/sda

2005-08-10 by Hans Blom
Why: To have as much as possible the same kernels at all nodes.
Where: On nodes "rembrandt1,2".
How:
In principle the guidelines are followed as described in Web100KernelInstallation and the kernel installation part from ChelsioDriver. Unfortunately "rembrandt1" and "rembrandt2" are different from the other Rembdrandt nodes in the sense that SATA happens to be there the first drive. Probably for this reason it is required that the 3w_9xxx driver is included in the kernel. The kernel has been build and installed with the following steps:

BIOS upgrade

2005-08-03 by Paola Grosso
Why: To (possibly) solve some stability problems seen during 10GE network tests
Where: On node "rembrandt3,4,5,8"
How:
Read note below. This is just the follow up of the previous work

BIOS upgrade

2005-07-25 by Bert Andree and Paola Grosso
Why: To (possibly) solve some stability problems seen during 10GE network tests
Where: On node "rembrandt1,2,6,7". "rembrandt3,4,5,8" will be done at a later stage when the Tier1 test are finished
How:

Install Rancid Tool

2005-07-20 by Hans Blom
Why: To store the configuration of the switches with version control.
Where: On node "rembrandt0".
How:
  1. Create a special user "rancid" to run the tool. The zsh shell had been used here because all shell and configuration scripts are Bourne shell oriented.
  2. Su to that user. Unpack the tool and install it below the home directory of the "rancid" user:
    $ su - rancid
    $ tar xfvz rancid-2.3.1.tar.gz
    $ cd rancid-2.3.1
    $ ./configure --prefix=$HOME/Rancid
    $ make install
  3. Make the following configuration changes in ~/Rancid/etc/rancid.conf:
    • Use the umask defined in this file, 027, also in CVS by addition of: CVSUMASK=027; export CVSUMASK.
    • Define the group switches to use by addition of: LIST_OF_GROUPS=uva-switches.
    • Because we intend to run the Rancid tool once per day in the place of the default once per hour, OLDTIME=4 is changed into OLDTIME=25. OLDTIME (in hours) should be grater than the sampling period.
  4. State changes to the uva-switches will be reported to the mail aliases rancid-uva-switches or rancid-admin-uva-switches. Therefore, these aliases have been defined in /etc/aliases. After that the command newaliases has been run.
  5. Edit ~/.cloginrc such that it containes per route configuration lines about the login procedure such as:
    add user {<router>} {<user>}
    add password {<router>}  {password}
    add method {<router>} {ssh}
    This file will be parced by tcl. Therefore, the {...} constructs are used to prevent tcl parsing it. Because ~/.cloginrc contains user/password information its should not be readable by others. At every switch there has been checked if the configuration is correct by running clogin <router..
  6. Run rancid-cvs to setup the CVS directries and files for the groups in LIST_OF_GROUPS.
  7. Define the routers in the uva-switches group by adding to ~/Rancid/var/uva-switches/router.db lines:
    <router>:<manufacturer>:up
    Unfortunately the passport switch is insufficient supported by the available script ~/Rancid/bin/brancid for baynet/nortel. Therefore, using the brancid as base we have defined a passport manufacturer in the following way:
    • Create a passport-rancid script from the brancid script:
      $ cd ~/Rancid/bin
      $ cp -i brancid passport-rancid
      In this script a.o. the following modifications have been performed:
      • Replace blogin by clogin because running blogin <router> was hanging wile clogin was successful.
      • Limit the commands to run at the router by show config only. Also no exit command becase that generates an expect EOF error.
      • Remove the time stamp on top of the configuration listening because otherwise CVS will change the revision after each run.
    • Link the passport manufacturer to the passport-rancid by additing the line
      elsif ($vendor =~ /^passport$/i) { exec('passport-rancid', $router); }
      at the end of the if-list on top of the ~/Rancid/bin/rancid-fe script. But first the original has been stored in /Rancid/bin/rancid-fe.orig.
  8. The command rancid-run is run manually to check if everything is working fine. The output will come in the most recent ~/Rancid/var/logs/uva-switches.*.
  9. Run the tool once a day at 04:02 (two minutes after the last Cricket collecting) by installing the crontab entry:
    2 4 * * *  $HOME/Bin/handle-rancid-run >$HOME/Log/handle-rancid-run.log 2>&1
    The script $HOME/Bin/handle-rancid-run handles the following actions.
    1. Run the rancid-run command.
    2. Remove the log files that are older than 48 days.
    3. Backup the CVS main directory ~/Rancid/var/CVS with an rsync command to the same user at "vangogh0" which is created there for that purpose. Also the CVS directory at "vangogh0" is identical there.

Change /space to Raid5

2005-06-07 by FreekDijkstra
Why: To have more reliable disk /space.
Where: On Rembrandt0 only (the rest remains RAID5)
How:
su -
df -h    # To determine the "/space" device.
cd /space

# Backup files (large!):
tar czf mirrors.tgz mirrors
tar czf user.tgz user
scp user.tgz freek@145.100.3.88:/space/user/freek/
scp mirrors.tgz freek@145.100.3.88:/space/user/freek/

cd
umount /space

tw_cli    # Give the following commands in the "tw_cli" shell.
show    # To determine the controler.
/c4 show    # To determine the unit and the # disks.
/c4/u0 del    # Delete the unit.
/c4 add type=raid5 disk=0-7    # Create the Raid5 units over all disks.
quit

apt-get install xfs
mkfs.xfs /dev/sdb    # Make a XFS file system.
mount /space
cd /space

# Restore files
scp freek@145.100.3.88:/space/user/freek/user.tgz ./
scp freek@145.100.3.88:/space/user/freek/mirrors.tgz ./
tar xfz user.tgz
tar xfz mirrors.tgz
rm user.tgz mirrors.tgz

Change /space to Raid0

2005-06-07 by Hans Blom
Why: To have a faster disk access to /space.
Where: On Rembrandt5 - 8 (the other nodes were already Raid0).
How:
su -
df -h    # To determine the "/space" device.
cd /space
tar cvf - user | gzip -c -v > /home/space_user.tar.gz    # Backup the user directories, "/home" is largest partition.
cd
umount /space

tw_cli    # Give the following commands in the "tw_cli" shell.
show    # To determine the controler.
/c4 show    # To determine the unit and the # disks.
/c4/u0 del    # Delete the unit.
/c4 add type=raid0 disk=0-10    # Create the Raid0 units at all disks.
quit

mkfs.xfs /dev/sdb    # Make a XFS file system.
mount /space
cd /space
tar xfvz /home/space_user.tar.gz    # Restore the user directories.
rm /home/space_user.tar.gz

Install Cogito and GIT

2005-05-23 by Hans Blom
Why: To be able to use pre-release kernel sources.
Where: On all Rembrandt nodes.
How:
cd /usr/local
tar xfvj <InstallDir>/cogito-0.10.tar.bz2
cd cogito-0.10
export PATH=$PATH\:`pwd`
make
cg-clone -s rsync://rsync.kernel.org/pub/scm/cogito/cogito.git
cg-branch-add pasky rsync://rsync.kernel.org/pub/scm/cogito/cogito.git

Add /usr/local/cogito-0.10 to the path in the various shell resource files.

Load the 3w-9xxx SATA driver module at boot time

2005-05-09 by Hans Blom
Why:
The RAID /space partition is not mounted during system boot, probably because the 3w-9xxx SATA driver module has been loaded into the kernel after the mount commands.
Where: On all Rembrandt nodes.
How:
Add the line 3w-9xxx at the end of /etc/modules. However, at the hosts where the 3w-9xxx SATA driver has been add to the kernel (currently "rembrandt1" and "rembrandt2"), add the driver name commented out and preceded by an informative message.

Create copy of system tcpdump usable for authorised

2005-04-29 by Hans Blom
Why: Users of the nettest group can now run this command.
Where: On all Rembrandt nodes.
How:
cp -pi /usr/sbin/tcpdump /usr/local/sbin
chgrp nettest /usr/local/sbin/tcpdump
chmod u+s,o-rx /usr/local/sbin/tcpdump

Created non-RAID /space/user/User directories

2005-04-22 by Hans Blom
Why: To be able to have local user directories when RAID is not available.
Where: On "rembrandt7"
How:
mkdir -p /home/space/user
ln -s /home/space/user /space/user
mkdir /space/user/User
chown User:Group /space/user/User
...

Modified bash and zsh startup files

2005-04-11 by Hans Blom
Why: To be able to set the user environment for non-interactive zsh shells.
Where: All Rembrandt nodes.
How: See the ShellStartupFiles page.

Fixed small bugs in Wikka code

2005-03-01 by FreekDijkstra
Why: because I was annoyed by these bugs ;-)
Where: on remote.science.uva.nl
How: see WikiCodeChanges page

Created various symbolic links to (automount) home directory roots such that all paths at all hosts do exist.

2005-02-16 by HansBlom
Why: Full home directory paths in user resource files will always be defined.
Where: All Rembrandt nodes.
How:

Backup the home directories from server "rembrandt0" in two cluster nodes.

2005-02-01 by HansBlom
Why: To have a backup at disk failures.
Where: Nodes "rembrandt0" (src) and "rembrandt1", "rembrandt2" (dest).
How:
The script rembrandt0:/root/Lib/Bin/backup_users makes for each user User a tar-gzip archive that will be created via SSH and a pipe at: rembrandt1-2:~backhome/Backup/REMBRANDT0/home/User.tar.gz (local home directories) and rembrandt1-2:~backhome/Backup/REMBRANDT0/home/vangogh0/User.tar.gz (NFS mounted home directories). The backup user "backhome" is a local user. The sub directory Backup is in fact a link to a local directory in the /space RAID such that there will be sufficient space. The archives of no longer existing home-directories at "rembrandt0" will not be automatically removed. The crontab entry rembrandt0:/etc/cron.d/backup_users takes care that the backup of the users will be repeated each Sunday morning.

Configured also NFS mounts at home directory server "rembrandt0".

2005-01-31 by Hans Blom
Why: To be able to NFS mount home directories also from "rembrandt0".
Where: Nodes "rembrandt1-8".
How:

Configured the home directory server.

2005-01-31 by Hans Blom
Why: To be able to NFS mount the home directories at Rembrandt and VanGogh.
Where: Node "rembrandt0".
How:

Configured the TCP wrapper.

2005-01-31 by Hans Blom
Why: To obtain more security for some servers, especially for the portmapper.
Where: All Rembrandt nodes.
How:
Give full access to sshd and for Globus in.ftpd and globus-gridftp-server. Deny access to all other services.

Change the home directory mount source at "vangogh0".

2005-01-31 by Hans Blom
Why: The home directories at "vangogh0" have been moved to that new location.
Where: All Rembrandt nodes.
How:
Mount the home directories at "vangogh0" now from /disk2/home/vangogh0 in the place of /disk2/home/nfs with the following steps:

Installed PHP package.

2005-01-26 by Hans Blom
Why: To be able to run PHP scripts.
Where: All "Rembrandt" nodes.
How:
apt-get [--dry-run] install php4-cgi php4-dev php4-pear phpdoc

Made bash user resources also suited for zsh.

2005-01-18 by Hans Blom
Why: To define proper settings for zsh users.
Where: All "Rembrandt" nodes.
How:

Synchronize packages

2005-01-10 by Freek Dijkstra
Why: Same packages installed on all nodes
How:
Optionally upgrade all nodes first:
/etc/cron.weekly/syncdebmirror on rembrandt0
apt-get update on all nodes
apt-get dist-upgrade on one or all nodes
get a list of packges of each node using:
ssh_rembrandt "dpkg --get-selections > \`uname -n\`"
Create a file with the desired packages, based on the current installation (using the output of dpkg --get-selections)
Update packages on all nodes:
apt-get update
dpkg --set-selections < myselection
apt-get -u dselect-upgrade

Created new changelog on rembrandt0

2005-01-10 by Freek Dijkstra
Why: New year, new order
How:
mv, cp :-)
In this file, start with newest entries (instead of at the bottom)

Synchronized /etc/apt/sources.list

2005-01-10 by Freek Dijkstra
Why: Now all nodes use debian mirror on rembrandt0
How:
rem_rembrandt_oper -filedistrib /etc/apt/sources.list
rem_rembrandt_oper -execute "apt-get update" rembrandt[0-8]

Change in rem_node_oper script

2005-01-10 by Freek Dijkstra
Why: -filedistrib didn't work
How:
near scp routine, changed -re "password" to -re "(P|p)assword"

Installed sudo package

2005-01-10 by Freek Dijkstra
Why: I like it
How:
on rembrandt[0-8]: apt-get install sudo
Added "%ii ALL=(ALL) ALL" to /etc/sudoers

Created globus user

2005-01-10 by Freek Dijkstra
Why:
globus user as a non-root account for installation and configuration of the globus software, just like on the VanGogh nodes
How:
On rembrandt[0-8]:
groupadd -g 55000
add_cluster_user -D -c 'Globus User' -g 55000 -s /bin/bash -u 55001 globus

Build & installed various software from source in /usr/local

2005-01-07 by HansBlom
Why: To give the Rembrandt nodes the same functionality as the VanGogh nodes.
How:
Using the archives stored at /usr/local/src from VanGogh node "vangogh1" as example. Follow the instructions from the README and INSTALL files in the packages top directories. Obtained installation problems and used non-standard commands have been discussed in the HOWTO /usr/local/doc/system/HOWTO/install_source_software.txt. Also the shell tcsh, bash shell startup files and the man config. files have been adjusted correspondingly. For bash only the environment for non-root users will be defined by sourcing the file /etc/bashrc.user only when the user ID is unequal zero. The "SABUL" package could not be build because its ./configure script did fail. Anyhow its successor UDT could be installed (but not without problems, see also the HOWTO mentioned above).

Added user accounts from VanGogh to Rembrandt

2005-01-06 by HansBlom
Why: Such that they will also be able to login at the Rembrandt cluster.
Where: Nodes "rembrandt[0-8]".
How:
At every node useradd has been run with the data listed in /etc/passwd at the VanGogh cluster. The shadow passwords listed in /etc/shadow have been copied with vipw -s. Users that have not used the VanGogh cluster for about a year have not been add to the Rembrandt cluster.

Synchronised the Debian packages from source "rembrandt0"

2005-01-04 by HansBlom
Why: Use at all nodes about the same Debian packages.
Where: Nodes "rembrandt[0-8]".
How:
At ever node there exixts a directory ~root/Handle_Packes containing a.o. two scripts:
First at node "rembrandt0" the command
./list_installed > installed_rembrandt0
has been run. Some packages only used at "rembrandt0" have been removed from this file. This adjusted file has been copied to the other nodes. At nodes "rembrandt[1-8]" the command
./list_installed > installed_`hostname -s`
has been run. Hereafter the packages are installed with the command
apt-get [--dry-run] install `list_unistalled`
when needed preceded by apt-get update.

Categories
CategoryLogs

There are no comments on this page. [Add comment]

Valid XHTML 1.0 Transitional :: Valid CSS :: Powered by Wikka Wakka Wiki 1.1.6.0
Page was generated in 0.2657 seconds