Difference between revisions of "Debian installation"

From Biowikifarm Metawiki
Jump to: navigation, search
(Related: Reconfiguring a copied virtual machine)
Line 151: Line 151:
 
==== Related: Reconfiguring a copied virtual machine ====
 
==== Related: Reconfiguring a copied virtual machine ====
  
Before starting a copied VM check the MAC adress in XEN to avoid running accidentially two servers under the same IP. A copied VM may use other interfaces instead of eth0, e.g., eth2. Thus first use ifconfig -a to find the name of the interface, then change interface name AND IP-number in network AND shorewall interface files. Also edit hostname and hosts, and restart shorewall:
+
Before starting a copied VM check the MAC adress in XEN to avoid running accidentially two servers under the same IP. A copied VM may or may not use other interfaces instead of eth0, e.g., eth2. Thus first use ifconfig -a to find the name of the interface, then change interface name AND IP-number in network AND shorewall interface files. Use the interfaces shown there. Also edit hostname and hosts, and restart shorewall. (Note: .57 is good, .58 bad...).
  
 
  sudo ifconfig -a | more
 
  sudo ifconfig -a | more

Revision as of 04:06, 24 August 2012

(Part of the Software documentation of biowikifarm.)

This is a record of the installation and later updates of a Debian system (originally in version 4, later upgraded, see "Debian upgrade 4 to 5"). The page is intended to help running or repeating the server setup. In parts it may be outdated by subsequent configuration changes and software updates. Some notes may be updated, others not. However, much of the fundamental setup will remain evident from this document.

We use XenServer Express (see Server virtualization) and originally installed the Xen template for Debian 4.0 (kernel modified for paravirtualization) with the standard setup. This is a documentation of the steps performed to get the Debian server working. The notes have been updated since, now reflecting a debian 6 setup.

See also: Biowikifarm server hardware.

Basics

The server can be reached at 160.45.63.55 with ssh enabled. Settings for ssh:

  • /etc/ssh/ssh_config default settings for ssh when used as client
  • /etc/ssh/sshd_config default settings for ssh as server
    • use nano /etc/ssh/sshd_config to set
      PermitRootLogin no
      PasswordAuthentication no
      UsePAM no
      = disallowing root access through ssh and limiting access to ssh pubkey infrastructure.
    • use nano /etc/pam.d/common-password to increase minimum password length:
      "password [success=1 default=ignore] pam_unix.so obscure sha512 min=11 max=30 nullok retry=3"
    • check lastlog to see last login of users

Current port and firewall situation: See Server ports and shorewall installation.

Memo on other IPs: 10 = base, 11, 41 = xencenter, and 56 = dev

Partitioning and formatting additional storage

NOTE: Info below refers first to older setup, newer setup is below

Xen will partition and format the first two hard drives (system and swap), but any additional, medium needs to be partitioned manually. The name of the third virtual hard disk under XEN is xvdc. Run:

fdisk /dev/xvdc
#  type n for new partition, p for primary, 1 for first partition, 
#  accept defaults for size, type w to write out new partition. 
# To format with ext3 filesystem:
mkfs.ext3 /dev/xvdc1
# repeat the above for a temp and backup disk (/dev/xvde + /dev/xvde1), then
# To label partitions:
e2label /dev/xvda1 root
e2label /dev/xvdc1 storage
e2label /dev/xvde1 dump

Note that the fourth hard disk (used for dispensible content like logs or backup duplicates) had been xvde, with xvdd being fixed to the DVD device.

To permanently mount, add create new folders in /mnt, then add the following line to "/etc/fstab" (nano /etc/fstab):

/dev/xvdc1      /mnt/storage    ext3    rw,noatime      0       0
/dev/xvde1      /mnt/dump       ext3    rw,noatime      0       0
# to mount temporarily use something like:  "mount /dev/xvdc1 /mnt/temp"

During Xen-Debian install, the swap partition had been changed from 512 to 1024, however, to actually use more than 512 one needs to issue:

mkswap /dev/xvdb1 -L swap
swapon /dev/xvdb1

Check with

cat /proc/swaps
free




NOTE: Here the updated information for debian6 on xen6. Swap is here in main.

Xen will partition and format the first hard drive (main = system and swap), but any additional, medium needs to be partitioned manually. The name of the 2./3 virtual hard disks under XEN is xvdb/xvdc. Run:

sudo fdisk /dev/xvdb
sudo fdisk /dev/xvdc
# type c and u to disable DOS-compatibility,
# type n for new partition, p for primary, 1 for first partition, 
#  accept defaults for size, type w to write out new partition. 
# To format with ext3 filesystem:
sudo mkfs.ext3 /dev/xvdb1
sudo mkfs.ext3 /dev/xvdc1
# To label partitions:
sudo e2label /dev/xvda1 root
sudo e2label /dev/xvdb1 storage
sudo e2label /dev/xvdc1 dump

Note that xvdd1 is the CD/DVD-ROM! It is under /media/cdrom0, not under /mnt/cdrom0!

Correct function of partitioning can be tested, e.g. using webmin (Hardware, Partitions on Local disks). To permanently mount, do

sudo nano /etc/fstab
#then edit:
/dev/xvdb1      /mnt/storage    ext3    rw,noatime      0       0
/dev/xvdc1      /mnt/dump       ext3    rw,noatime      0       0
# to mount temporarily use something like:  "mount /dev/xvdc1 /mnt/temp"

Check swap with

cat /proc/swaps
free

(Result: /dev/xvda5!)

Resizing virtual disks

  1. The disk needs to be deactivated in XenCenter before resizing.
  2. Resizing the virtual storage unit does not change the size of the partition under Linux!
  3. Method 1 (manual): if the disk contains only relatively easily migrated data (no dev/ports, etc.), copy all content to another disk using cp -pr (preserving owner, permission, recursive) or tar -p -s. Use fdisk as above, delete and recreate partition, format and label disk. Copy all back, delete temporary copy.
  4. Method 2 (better, tip Manol, not yet tested):
    • install parted: apt-get install parted or gparted (graphical frontend)
    • unmount partition and check with parted
    • OR: install apt-get install dump (dump - 4.4bsd dump and restore for ext2 filesystems) and use this to store and

Network, Hostname and IP address

First installation

Hostnames are in several places, especially important when renaming a server:

sudo nano /etc/hostname # entered: biowikifarm (old) biowiki (new) 
sudo nano /etc/mailname # entered: biowikifarm.net
sudo nano /etc/hosts    # 160.45.63.55 biowikifarm.net biowikifarm
sudo nano /etc/hosts    # 160.45.63.58 biowikifarm.net biowiki

Open "nano /etc/network/interfaces" and enter: NEW:

# The loopback network interface
auto lo
iface lo inet loopback
# The primary network interface
allow-hotplug eth0
iface eth0 inet static
       address 160.45.63.58
       netmask 255.255.255.0
       network 160.45.63.0
       broadcast 160.45.63.255
       gateway 160.45.63.1
       # dns-* options are implemented by the resolvconf package, if installed
       dns-nameservers 160.45.8.8 160.45.10.12
       dns-search bgbm.fu-berlin.de


# The following lines are desirable for IPv6 capable hosts
::1     ip6-localhost ip6-loopback
fe00::0 ip6-localnet
ff00::0 ip6-mcastprefix
ff02::1 ip6-allnodes
ff02::2 ip6-allrouters
ff02::3 ip6-allhosts
(Note: 160.45.8.8 and 160.45.10.12 are the FU DNS)

Reboot now (first command) and run the following commands (the two hostname commands verify that the hostname setting is ok):

sudo reboot
hostname     # verify
hostname -f  # verify
sudo apt-get update; sudo apt-get upgrade;
sudo apt-get dist-upgrade
sudo apt-get install ssh sudo

It may be good to repeat this, some updates came only in second round. Finally, we may want to do some cleanup:

sudo apt-get autoclean

After doing this, the ssh fingerprints may have to be changed on all machines trying to access the new Debian by ssh (security issues fixed). Remove the keys for the previously opened IPs in the following file:

nano /root/.ssh/known_hosts

Related: Reconfiguring a copied virtual machine

Before starting a copied VM check the MAC adress in XEN to avoid running accidentially two servers under the same IP. A copied VM may or may not use other interfaces instead of eth0, e.g., eth2. Thus first use ifconfig -a to find the name of the interface, then change interface name AND IP-number in network AND shorewall interface files. Use the interfaces shown there. Also edit hostname and hosts, and restart shorewall. (Note: .57 is good, .58 bad...).

sudo ifconfig -a | more
sudo nano /etc/network/interfaces # interface and IP
sudo nano /etc/shorewall/interfaces # (interface at the bottom, check, update if necessary)
sudo nano /etc/hostname
sudo nano /etc/hosts
sudo /etc/init.d/shorewall restart

(Note: if due to a misconfiguration the shorewall continuously displays blocking messages on the console, try Alt-F2, user name and password. Type blindly, despite repeated shorewall messages. "shorewall stop" will stop shorewall (type blindly).)

sudoers

sudo nano /etc/sudoers
# then check wether (this is the default of debian6 or xen template):
%sudo   ALL=(ALL) ALL

This lets all users of group sudo run any command as root without knowing the root password.

Adding repositories to package manager

It is possible to add further repositories to the package manager by editing the repository file: "nano /etc/apt/sources.list". One important addition is the backports (needed for OpenID further down), and we also expanded the scope from "main" to "main non-free contrib":

deb http://ftp.debian.org/debian/ etch main non-free contrib
deb http://www.backports.org/debian etch-backports main 
# Necessary for webmin (sarge is still the correct version for squeeze!!):
deb http://download.webmin.com/download/repository sarge contrib
# added for oggvideotools (June 13 2010)
deb http://www.debian-multimedia.org lenny main

LONGER WORKING: To install keys: wget -O - http://backports.org/debian/archive.key | apt-key add -  ; sudo apt-get install debian-backports-keyring; --- The last command adds a gpg security key for the backport repository. (Documentation of further testing: We also tested the Google repository as an example. Using root console, we first had to get a pgp key (note: the original google instructions requested using https, but only http worked): wget -q http://dl-ssl.google.com/linux/linux_signing_key.pub -O- | apt-key add -, and add deb http://dl.google.com/linux/deb/ stable non-free.)

# more on clamav and rkhunter below, but install early:
sudo apt-get install clamav rkhunter

JAVA

Debian does not normally install Sun Java, but a slower "free Java". It is possible to install sun software after modifying the repository sources to include non-free software and backports (see here for more information: http://www.debian.org/doc/manuals/debian-java-faq/ch6.html):

sudo apt-get install sun-java6-jre

The install location is /usr/lib/jvm/java-6-sun/. Sun recommends to update the Debian "alternatives system" to have Sun's tools as the default:

sudo update-java-alternatives -s java-6-sun

Then edit

sudo nano /etc/profile

# and add at the top, before first "if":
export JAVA_HOME=/usr/lib/jvm/java-6-sun
export JRE_HOME=/usr/lib/jvm/java-6-sun/jre
export CATALINA_HOME=/usr/share/fedora/tomcat
export FEDORA_HOME=/usr/share/fedora

# and add further down, after last "fi":
PATH=$JAVA_HOME/bin:$FEDORA_HOME/server/bin:$FEDORA_HOME/client/bin:$PATH
export PATH
umask 022
HISTFILE=~/.bash_history
HISTSIZE=10000
HISTFILESIZE=999999
readonly HISTFILE
readonly HISTSIZE
readonly HISTFILESIZE
export HISTFILE HISTSIZE HISTFILESIZE

(The last line before "export PATH"; this export is probably only necessary (e.g. for Tomcat, see below) if an earlier version of Java is still running and is located in a path that is looked up first by the system.)

Log out and log back in again (this may also be another user) and then confirm that these settings are in effect by:

echo $JAVA_HOME
echo $JRE_HOME
echo $PATH

MySQL, Apache PHP, etc.

General installations

sudo apt-get install gcc cpp libtool binutils make autoconf automake1.9 flex tree locate 
sudo apt-get install zip unzip rar unrar p7zip p7zip-full
sudo apt-get install librsvg2-2 librsvg2-bin

The librsvg2-2/librsvg2-bin is an svg-renderer used by MediaWiki (requiring some extra work in settings files though). In our installation, it was already installed with a previous package.

To use MediaWiki with high traffic, a memcached memory cache should be installed prior to installing MediaWiki:

# Presently not used, using PHP-APC instead:
# apt-get install memcached php5-memcache

Without further configuration, a single memcached will be active on default port 11211, limited to localhost (testing from outside with telnet 160.45.63.55 11211 confirmed this), using up to 64 MB of your RAM. This can not be changed by editing /etc/init.d/memcached, but /etc/memcached.config must be modified. We increased size: -m 128 = 128 MByte, user: -u www-data (run under same user as Apache) and -l 127.0.0.1 = listen only on local loopback interface.

Note: At the time of this writing, a fairly old version of memcached (1.1.12) is installed by apt-get of Debian 4 and No backports are available.

Update: at present, memcached has been de-installed because PHP-APC is used as object cache instead (single-machine setup).

MySQL

Installation

 # debian5: sudo apt-get install mysql-server mysql-client libmysqlclient15-dev
 # on debian6 we used:
 sudo apt-get install mysql-server mysql-client libmysqlclient-dev

 sudo nano /etc/mysql/my.cnf
 # making many changes, use a merge with a current config file
 /etc/init.d/mysql restart

Under Debian 6, We experienced a broken apt-get after an "apt-get remove mysql-server mysql-client libmysqlclient-dev" and then "apt-get install --reinstall mysql-server mysql-client libmysqlclient-dev" and removing the content of /var/lib/mysql (to start fresh over, after the first dump import failed). In the end, removing further parts or mysql AND manually installing mysql-server-core-5.1 fixed the broken installation situation!

Console commands

Test with netstat (debian seems to report 8088 as omniorb) and set a password for root (otherwise anybody can access your MySQL database!):

 netstat -tap
 # -u = login as this user
 mysqladmin -u root password (ThePassword)

New users can be created and granted privileges inside the mysql console:

 # enter the mysql console (asks for passwort):
 mysql -u root -p
 -- within the console (mysql> ):
 grant all privileges on *.* to 'username'@'localhost' identified by 'PASSWORD' with grant option;
 grant all privileges on *.* to 'username'@'%' identified by 'PASSWORD' with grant option;
 -- to exit the mysql console type:
 exit

Notes: a) grant will create a user if it does not exist yet. b) username, hostname, PASSWORD must be quoted with '. c) the '%' allows the user to log in from any other host except localhost; an explicit grant for localhost is thus required in addition to the '%'.


Changing password for existing users:

 # enter the mysql console (asks for passwort):
 mysql -u root -p
 -- within the console (mysql> ):
 set password = PASSWORD('NEW-PASSWORD-HERE');
 -- Alternative, for a different user:
 set password FOR 'name'@'localhost' = PASSWORD('NEW-PASSWORD-HERE');

Some other mysql console commands are:

 show databases;
 use mysql; Select * from user;
 drop user root@OudemansD;

Important note: entering mysql command line with "mysql -u wikiuser -p" will ask for password; if you want to include the password (e.g. when using pipe into) do not add a blank between p and the password. Also, when using the -h host option, only -hlocalhost, but not -hlocalhost:8088 will work! The host can usually be omitted. See the following examples:

 mysql -uwikiuser -pTHEPASSWORD
 mysql -hlocalhost -u wikiuser -pTHEPASSWORD < openid_table.sql

Copying or importing a MySQL database

Unlike MS SQL Server, you cannot reliably move a MySQL database by detaching, moving, and attaching binary files in the MySQL directory. You must dump the mysql database to a sql/text file using the mysqldump command. The SQL commands can then re-create the tables and contents. To use the file, start the mysql command interface, create and use the database and give the command "source your_filename" which reads and executes the commands from the dump file.

Over the web, myphpadmim offers backup (which works) and restore (which does not work for any reasonably sized database, despite uploading as zip file, we are getting out of memory errors). So we have to use command line programming. Note that the dump created by myphpadmin includes a database create statement; to copy to a different database name you need to modify this (extremely long lines may become corrupted in some editors!); the mysqldump does not contain database statements and is easier to reuse.

To dump a database:

 mysqldump -u root -p wikidb > /var/backups/wikis/wikidb.sql

This writes the database into a file called sample.sql. To create or import it into a new database called xxx, creating tables with their inserts that were present in the former wikidb, use:

 mysql -u root -pMySecretPW
 -- DROP DATABASE IF EXIST `xxx`;
 CREATE DATABASE `xxx`;
 -- for a wiki:
 CREATE DATABASE metawiki DEFAULT CHARACTER SET utf8 COLLATE utf8_general_ci;
 USE `xxx`;
 -- load from a file:
 SOURCE /var/backups/wikis/wikidb.sql;

You can also copy directly a database from one server to another or the same server. The second database must have been created however,

 mysql -u root -p
 CREATE DATABASE metawiki DEFAULT CHARACTER SET utf8 COLLATE utf8_general_ci;
 exit;
 mysqldump -u root -p(ThePassword) --opt wikidb | mysql -u root -p(ThePassword) --host=localhost -C metawiki

Here wikidb is copied to "localhost" (–C tells mysqldump to use data compression if both servers support it). "localhost" can be replaced by the IP address of any local or remote computer.

Note: a wiki-backup script has been started in /var/backups/scripts

Execute sql from command line:

 mysql -u root -p --database=orowiki < /var/www/v-orowiki/w/media/orowiki.sql


Performance tuning and unicode:

In /etc/mysql/my.cnf we changed:

key_buffer      = 64M  # was 16M
# utf8
init-connect='SET NAMES utf8'
character-set-server=utf8
collation-server=utf8_unicode_ci

We include three-letter words in the full text index:

ft_min_word_len=3

(Note: after changing this, run "REPAIR TABLE searchindex QUICK;" in mysql/myadmin on each affected database.)

Updating user passwords

mysqladmin -u root -p'theOLDpassword' password 'theNEWpassword'

Version upgrading/moving all databases: See MySQL Backup and Restore


Apache

sudo apt-get install apache2 apache2-doc apache2-mpm-prefork apache2-utils libexpat1 ssl-cert libapache2-mod-wsgi
sudo nano /etc/apache2/mods-available/dir.conf
# add more values to DirectoryIndex:
#    index.html index.htm index.shtml index.cgi index.php index.php3 index.pl index.xhtml
nano /etc/apache2/ports.conf
# add ssl port to the existing "Listen 80" on new line: Listen 443

Now we have to enable some Apache modules (SSL, rewrite, suexec, and include) and execute the last command restart:

sudo a2enmod ssl
sudo a2enmod rewrite
sudo a2enmod suexec
sudo a2enmod include
sudo /etc/init.d/apache2 force-reload 

Apache config is in /etc/apache2/apache2.conf, the virtual host configurations are included from: /etc/apache2/sites-enabled/ which are links pointing to /etc/apache2/sites-available/. The command sudo a2ensite name enables an available site.

By default, the www-root (DocumentRoot) is /var/www/.

To test the configuration, use sudo /usr/sbin/apache2 -t

SSL: We use a self-signed certificate, using the OpenSSL command:

sudo openssl req -new -x509 -days 2190 -nodes -out /etc/ssl/certs/ssl-cert-XXXX.pem -keyout /etc/ssl/certs/ssl-cert-XXXX.key

Note: Do not leave the "personal name" or "common name" empty. Apache will not complain, but subversion will complain about a missing commonname attribute ("Server certificate was missing commonName attribute in subject name").

To enable SSL, link in sites-enabled to /etc/apache2/sites-available/default-ssl, modify default-ssl by pointing to the new pem and key files and enable the directive "NameVirtualHost *:443".

Restart:
sudo /etc/init.d/apache2 reload
(Note: forgetting the sudo, will result in cryptic messages like: "httpd not running, trying to start / (13)Permission denied: make_sock: could not bind to address [::]:80 / (13)Permission denied: make_sock: could not bind to address 0.0.0.0:80 / no listening sockets available, shutting down / Unable to open logs".)

Logrotation: sudo nano /etc/logrotate.d/apache2 - change any settings, e.g. daily or weekly, "create 640" to "create 660". logrotate has not problems with softlinks, but if moving the path, rights have to be observed. Test with sudo logrotate --debug /etc/logrotate.d/apache2 &> logrotate_debug_log.txt since logrotate writes to standard error, not standard out!

Subversion

apt-get install subversion libapache2-svn
# (note: modules are automatically enabled, no a2enmod necessary)
  • Tested with MediaWiki svn download (see below)
  • Potential additional instructions (not used): Configuring Subversion

Special commands:

svn update -rXXXX # update to specific revision (XXXX being an integer revision number)
svn revert filename # revert a modified file to its original state, so future updates will refresh the file again
cd /usr/share/mediawiki; svn status phase3 # show locally modified files (potentially no longer updated)

Later, a local repository was installed, see Configuring the subversion repository

PHP

Instructions for PHPMyAdmin: http://www.debianhelp.co.uk/phpmyadmin.htm, downloading not necessary, apt will do this.

sudo apt-get install libapache2-mod-php5 php5 php5-common php5-curl php5-dev php5-gd php5-idn php-pear php-mail php-net-smtp php5-imagick php5-imap php5-json php5-mcrypt php5-mhash php5-ming php5-mysql php5-ps php5-pspell php5-recode php5-snmp php5-sqlite php5-tidy php5-xmlrpc php5-xsl php5-intl
sudo apt-get install phpmyadmin

Note: Originally the installation included php5-idn as well. However, the two packages intl/idn seem to conflict. "apt-get install php5-intl php5-idn" results in: "The following packages have unmet dependencies: php5-intl : Conflicts: php5-idn". With only idn installed, we got the error "'/usr/lib/php5/20090626/intl.so' - /usr/lib/php5/20090626/intl.so: cannot open shared object file: No such file or directory". We have thus executed: "apt-get remove php5-idn; apt-get install php5-intl;"


Enable php after installing:

sudo a2enmod php5

This creates symbolic links from /etc/apache2/mods-available/php5.conf and /etc/apache2/mods-available/php5.load to /etc/apache2/mods-enabled.

As long as all php runs on a single server, APC is better than memcached. Recommended reading: http://www.mediawiki.org/wiki/User:Robchurch/Performance_tuning it is recommended to use:

sudo apt-get install php-apc

Check phpinfo.php, section apc for success. Note: Initially we did not configure the APC user cache for the mediawiki object cache ($wgMainCacheType = CACHE_ACCEL), but use memcached for this ($wgMainCacheType = CACHE_MEMCACHED) to simplify transition to multiserver setup without APC. However, we have since moved the entire cache action to memcached.

PECL extensions

NOTE: The following did not work as described on the debian 6 installation. It may be that it is unnecessary, since we could install php5-intl there. Else reinvestigate!

See http://www.jejik.com/articles/2008/07/howto_build_and_install_the_intl_pecl_extension_for_php5_in_debian/

# check whether gpg keys available:
sudo  gpg --list-secret-keys
# if fail, then: sudo  gpg --gen-key
## libicu38 no longer works: sudo apt-get install php5-dev dh-make-php fakeroot libicu38 libicu-dev xsltproc
sudo apt-get install php5-dev dh-make-php fakeroot libicu-dev xsltproc
# download, extract, build:
cd ~/php-intl-1.1.2
sudo dh-make-pecl --depends libicu38 --build-depends libicu-dev --only 5 --maintainer 'YOUR_GPGName_and_EMAIL'  intl 
cd php-intl-1.1.2
dpkg-buildpackage -rfakeroot

(was build and installed by manol)

php.ini

For MediaWiki, uploading must be enabled. The php.ini file used by apache is at /etc/php5/apache2/php.ini and already contains file_uploads = On and register_globals was already Off. However several settings were changed:

; Maximum execution time of each script, in seconds
; http://php.net/max-execution-time
; DEFAULT: max_execution_time = 30
; Note: 240 still fails VERY large mediawiki identification keys
max_execution_time = 450
max_input_time = 90       ; Maximum amount of time in seconds each script may spend parsing request data
; Maximum amount of memory a script may consume (128MB)
; http://php.net/memory-limit
memory_limit = 260M
; Default was 128M. Maximum amount of memory a script may consume. 2 MB long mediawiki pages may consume 40M
; Before June 2012 we used: memory_limit = 300M  - now slightly reduced:
; NOTE GH: change suhosin.memory_limit in /etc/php5/conf.d/suhosin.in to same value!
upload_max_filesize = 50M
; DEFAULT !!! was just 2M!
; ALSO SET post_max_size = 52 ABOVE!
; This affects both form data and file uploads. This value must be larger than upload_max_filesize!
post_max_size = 52M 
; was 5, affects file upload and must be larger fileupload!
allow_call_time_pass_reference = Off

The ini file for command-line scripts is at /etc/php5/cli/php.ini; this is separate from the one used for the apache php module.

Syntax errors: The log files are relatively silence if a syntax error exists in php.ini. All values will be used with their defaults; thus this situation can be detected by checking Special:Upload whether it displays an upload limit of 2 MB. To manually check php.ini for syntax errors use:

php --info --php-ini /etc/php5/apache2/php.ini

Important for inserting utf-8 text into mysql:

Although a mysql table may be set to CHARACTER SET utf8, the mysql client's standard charset may still be Latin1. So when inserting utf-8 data with PHP, the client's charset may have to be changed to utf-8 after creating the connection:

$conn = mysql_connect($MySQL_Host, $MySQL_User,$MySQL_Passwd);
mysql_query("SET NAMES utf8", $conn);

suhosin.in

Change suhosin.memory_limit in /etc/php5/conf.d/suhosin.in to same value!

; DEFAULT WAS: suhosin.memory_limit = 0, now same value as in php.ini:
suhosin.memory_limit = 260M

apc.ini

This is in /etc/php5/conf.d/apc.ini

The status can be viewed by looking at apc_1417869461139.php in the www root.

Our content is (size increased from 30 to 60, especially for mediawiki user-cache usage):

extension=apc.so
# apc.shm_segments=3
apc.enabled=true
# default: apc.shm_size=30 # i.e. 30 MB
apc.shm_size=60

PHPMyAdmin

To enable phpmyadmin, edit: /etc/apache2/apache2.conf and add at end of file:

# Enable PHPMyAdmin
Include /etc/phpmyadmin/apache.conf

Changes in /etc/phpmyadmin/apache.conf for making Suhosin better compatible with phpmyadmin

# Changes recommended for suhosin:
   php_value suhosin.request.max_vars 2048
   php_value suhosin.request.max_value_length 1000000
   php_value suhosin.request.max_array_index_length 256
   php_value suhosin.request.max_totalname_length 8192
   php_value suhosin.post.max_vars 2048
   php_value suhosin.post.max_array_index_length 256
   php_value suhosin.post.max_totalname_length 8192
   php_value suhosin.post.max_value_length 1000000
   php_flag suhosin.sql.bailout_on_error Off
   php_value suhosin.log.file 0
   php_value suhosin.log.phpscript 0
   php_flag suhosin.log.phpscript.is_safe Off
   php_value suhosin.log.sapi 0
   php_value suhosin.log.script 0
   php_flag suhosin.log.use-x-forwarded-for Off


(Test phpmyadmin with "http://biowikifarm.net/phpmyadmin")


Django installation

See Django-Installation. For Artenquiz at least django Version 1.3 necessary!


Mediawiki

Mediawiki installation information is available separately: Mediawiki_installation


Tomcat

Important: to get Tomcat5.5 running after installing, it is necessary that the environment variables JAVA_HOME and JRE_HOME are set correctly and that JDK 5 or higher is used.

# NOTE: native tomcat no longer used:
# apt-get install tomcat5.5 tomcat5.5-admin tomcat5.5-webapps
apt-get install libapache2-mod-jk 

In /etc/default/tomcat5.5 remove comment markers from TOMCAT_USER=tomcat55 (default), or another user, and do the same for JAVA_HOME with the correct path (peferably a symbolic link to current version, e.g. JAVA_HOME=/usr/lib/jvm/java-6-sun). The other options only need to be uncommented if you want to override the defaults (which are set in /etc/init.d/tomcat5.5).

Edit /etc/profile (as user root), add:

## For debian standard: export CATALINA_HOME=/usr/share/tomcat5.5
## Instead, for fedora-tomcat:
export CATALINA_HOME=/usr/share/fedora-3.1/tomcat

To make Tomcat work with Apache edit /etc/libapache2-mod-jk/workers.properties and set the two lines (using your correct paths):

workers.tomcat_home=/usr/share/fedora-3.1/tomcat
workers.java_home=/usr/local/jdk

At the end of /etc/apache2/apache2.conf add:

# Enable libapache2-mod-jk
Include /usr/share/doc/libapache2-mod-jk/httpd_example_apache2.conf

--- Presently, the tomcat used is the one supplied with Fedora, inside the Fedora folder. The main tomcat access is fedora.keytonature.net. However, Lia Veja has installed a secondary point

 http://species-id.net/services/

L. Veja writes "it needs some edits in web.xml file for every application you will put there".

Restart Apache and Tomcat.

/etc/init.d/apache2 restart
/etc/init.d/tomcat5.5 restart

Fedora

See FEDORA Installation, FEDORA Batch Import

For manual Fedora 3.1 start method see section further below under "Stop services / demons"

Note: Fedora does not install the log folder in the log path of debian. GSearch logs become quickly very large (Gigabytes), so it is important to move them. The following code copies existing logs (assuming /var/log/fedora/ already created) and creates symbolic links. THIS NEEDS TO BE REPEATED WITH EVERY UPDATE OF FEDORA TO A NEW VERSION!

cp -pr /usr/share/fedora/server/logs/* /var/log/fedora/ # destination may be /mnt/dump/var/log/fedora
/usr/share/fedora/tomcat/bin/shutdown.sh
rm -r /usr/share/fedora/server/logs
ln -s /mnt/dump/var/log/fedora /usr/share/fedora/server/logs
# Old: /usr/share/fedora/tomcat/bin/startup.sh
# New, our own script:
/etc/init.d/fedora start


Note 2009-09:

http://fedora.keytonature.net/x/x.jpg

links to /var/lib/tomcat5.5/webapps/x/x.jpg with /usr/share/tomcat5.5/webapps/ being a link to /var/lib/tomcat5.5/webapps/. However the root of http://fedora.keytonature.net/ links to webapps/ROOT, but on 2009-09 was misconfigured to link to fedora-3.1/tomcat/webapps... folder.

Fedora Rebuild Index

In several situations it may be necessary to rebuild the Fedora indices:

  • Recovering from inconsistencies and/or corruption of the indexes.
  • Upgrading from a previous version of Fedora when the SQL database or Resource Index changed significantly between releases
  • Migrating from one SQL database product to another in an existing Fedora installation. Such migration can be done by modifying /usr/share/fedora-3.XXX/server/config/fedora.fcfg to point to a properly-configured <datastore..>

The batch service fedora-rebuild.sh must be started as follows:

* Stop the Fedora server (if using Fedora-Tomcat, this can be done with the /usr/share/fedora/tomcat/bin ./shutdown.sh command)
* Run /usr/share/fedora/server/bin ./fedora-rebuild.sh 
* Select option "1" in order to rebuild indexes
* Confirm this procedure with "1" 
* Restart the Fedora server (if using the Fedora-Tomcat, this can be done with the /usr/share/fedora/tomcat/bin ./startup.sh command)

See also: Fedora Commons documentation.

Webmin

Webmin is a web-based system configuration tool for Linux. With it you can configure many operating system internals, such as users, disk quotas, services, configuration files etc., as well as modify and control many open source apps, such as the Apache HTTP Server, PHP, MySQL etc.


To install, first /etc/apt/sources.list needs to be updated (see above). Then:

sudo apt-get update; sudo apt-get install webmin

You will be asked if you wish to install the package without verification. Select yes. To configure a different port:

sudo nano /etc/webmin/miniserv.conf
# change port AND listen options from 10000 to 18010 and restart:
sudo /etc/init.d/webmin restart

Only users with sudo rights can login with their usual password. Help for the standard modules used by Webmin can be found here: http://doxfer.com/Webmin/Modules. A wiki page for webmin can be found here: http://doxfer.com/Webmin

To increase security read this and:

  • in "Webmin: Webmin Configuration: Authentication" "set "Block hosts with more than 2 failed logins for 60 seconds; Block users with more than 6 failed logins for 12000 seconds" and select "Log blocked hosts, logins and authentication failures to syslog" and "Enable session authentication: Autologout after 10 Minutes" and disable "Offer to remember login permanently?". Rest remaining on defaults.
  • Create new certificates, then in "Webmin: Webmin Configuration: SSL Encryption" set: "Private key file: .../ssl-cert-XXX-webmin.key" and "Certificate file: Separate file: .../ssl-cert-XXX-webmin.key" and "Redirect non-SSL requests to SSL mode? Yes".

mc (midnight commander) und jed (editor)

Midnight commander is the Linux equivalent of the Norton Commander. Anyone familiar with Norton Commander should feel right at home. Very useful for browsing through the file system. It has an integrated FTP client, editor and file viewer, and supports the use of a mouse inside an ssh window! Similarly, jed is an editor that supports the use of mouse in ssh. Install both with

sudo apt-get install mc jed

No configuration necessary. Run with

mc
jed

cronjobs

DEPRECATED ON BIOWIKIFARM: to install as root user-specific cronjob (copy see nano /var/spool/cron/crontabs/root, but not editable there!), use:

sudo crontab -u root -e  # edit roots cron jobs

This uses jed as editor. Format at start of each line is: minutes, hour, day, etc. One can use * for every (hour, day, month, year). Output is normally sent by email. To suppress this redirect output to ">/dev/null 2&>1". Always add a comment after each line: # explain what this does and who installed it.

PREFERRED:

sudo nano /etc/crontab

This is directly editable, allows jobs for different users (www-data, root), and is more transparent to manage.

Security, virus protection, rootkit hunter

rootkit hunter:

# sudo apt-get install rkhunter
# OR, consider to install from backport, here squeeze backport a newer version:
sudo apt-get -t squeeze-backports install rkhunter
sudo rkhunter --update
sudo rkhunter --checkall

clamav:

sudo apt-get install clamav
/etc/init.d/clamav-freshclam start
# or: stop, restart, etc.

rest not sure. clamav, clamscan etc. don't work. Was not asking for configuration. -> ?

http://www.debianadmin.com/clamav-installation-and-configuration.html looks good, but did not help

General: use on files or folders to change group and rights:

sudo chgrp -R mw-extension-writers crossdomain*; sudo chmod -R 664 crossdomain*;
# or if containing folders:
sudo chgrp -R mw-extension-writers /var/www/tools/mediaIBIS; sudo chmod -R 775 /var/www/tools/mediaIBIS

Dropbox

Dropbox is installed for a specific user and with manual daemon startup. We created a special user dropbox for this task. dropbox is member of www-data, and www-data member of dropbox group. Following: https://www.dropbox.com/install?os=lnx

sudo su dropbox
cd ~ && wget -O - http://www.dropbox.com/download?plat=lnx.x86_64 | tar xzf -
wget https://www.dropbox.com/download?dl=packages/dropbox.py
~/.dropbox-dist/dropboxd

It will then ask to visit a URL from your desktop-browser, to confirm linking the new dropbox with an existing account. In the dropbox web page, go on top to account, then below the tab "My Computers". The procedure is a bit tough, since you have to copy the long link without stopping the command on the command line. For putty use right-click mouse, not Ctrl-C. The mouse-click will copy and paste, looking wrong, but the commandline process keeps running. The process was successfull if the command line on the server issues "Client successfully linked, Welcome (YourName)".

Added a brief shell script "run_dropbox.sh" with:

#!/bin/bash
~/.dropbox-dist/dropboxd&
ps

to help memorize Dropbox daemon startup.

Link the Dropbox folder inside dropbox user home into wiki for ease of importing:

cd /var/www/v-species/o; sudo ln -s /home/dropbox/Dropbox "Dropbox"

Tips for simple access of Debian file from remote Windows

Install WinSCP, which will allow to browse the Debian server almost like a Windows explorer (or a graphical Midnight commander). It allows to easily and securely (using sftp, we do not use ftp) create folders, copy files within the server, or upload/download files to your machine.

WinSCP also very nicely integrates with PuTTY, a ssh client. To make this as simple as possible, do the following:

  • You can either
    • install PuTTY (copy all files from the zip with all executables and help) into the Putty subfolder of WinSCP
    • Or (perhaps already) have installed it anywhere else
  • In any event go in WinSCP to the menu: Options/Preferences, then in the Dialog box: integration/applications and check the path. It should be something like "%PROGRAMFILES%\PuTTY\putty.exe" if putty has been installed in that folder.
  • In the same dialog, turn on Remember... and Automatically...
  • Now whenever you open WinSCP, a second window with ssh will automatically be opened.
  • To change the display options in PuTTY, go to the Window menu (top left), Change settings (e. g. Window, Color, Use System colors will use black-on-white text if you have set you Windows command window that way). Under Window increase Columns (e.g. to 120), Rows (e.g. to 63), and scrollback (e.g. to 900) to suit your taste. On "Session" itself select "Default" and click "Save" to preserve your changes.

Reversely, adding NTFS support to Debian: In our case this was desirable to add read/write support for NTFS-Formatted USB-disks. Provided backports are installed for apt-get (as instructed above), one can install:

aptitude install ntfs-3g

To temporarily mount from command line use (assuming the usb-disk is xvdf, check in Xen Center)

mount -t ntfs-3g /dev/xvdf1 /mnt/usb-disk
# (if disk is fat32, use: mount -t auto /dev/xvdf1 /mnt/usb-disk)

To mount the NTFS during boot add the following to the end of the /etc/fstab file:

/dev/xvdf1 /mnt/usb-disk ntfs-3g defaults 0 0

See ntfs-3g manual for further information. Some note mention that "To mount files with non-ASCII characters one may have to give the option -o locale=XXX to the mount options of ntfs-3g; see http://ntfs-3g.org/support.html#locale for further information." We did not do so far, using UTF-8 seems to work.


Moving files to a new server directly

Instead of backup/restore it is possible to go, on source machine, to the folder under which the subfolders shall be moved, then tar, ssh transfer, unpack. Caveat: on the destination machine the login MUST be root, else the user and group information will be lost. Temporarily change sudo nano /etc/ssh/sshd_config to permit root login!

cd /
# as root, being on old server:
sudo tar -czf - --numeric-owner root home var | ssh root@160.45.63.55 tar -xzf - --same-owner -C /mnt/dump/_oldroot_TEMP
sudo tar -czf - --numeric-owner var | ssh root@160.45.63.55 tar -xzf - --same-owner -C /mnt/storage/oldroot
cd /mnt/storage
sudo tar -czf - --numeric-owner * | ssh root@160.45.63.55 tar -xzf - --same-owner -C /mnt/storage
cd /mnt/dump
sudo tar -czf - --numeric-owner * | ssh root@160.45.63.55 tar -xzf - --same-owner -C /mnt/dump
cd /usr/share
sudo tar -czf - --numeric-owner mediaw* | ssh root@160.45.63.55 tar -xzf - --same-owner -C /mnt
sudo tar -czf - --numeric-owner fedora* Fedora* | ssh root@160.45.63.55 tar -xzf - --same-owner -C /mnt

Note: when moving, the user and group ID should be aligned, see: http://it.toolbox.com/blogs/locutus/how-to-change-a-users-uid-and-gid-26368