The idea explored here is to separate your data physically from the computer that you are using them on. This has the following advantages: (1) you are quite safe from theft because you can keep carrying your data with you always, even when having to let your notebook at home or in a hotel because it is too heavy; (2) you can use your data on any computer (like a friend's computer when travelling, an Internet cafe computer etc.); (3) you can hand over data items quickly and easily because every computer has a USB interface.

You have to decide if the this idea is worth its benefits for you, as all solutions below are much slower and more expensive than a bigger standard SATA III SSD (as of 2014-06, about 100 EUR new incl. shipment for a 256 GB SSD, with data transfer rates up to 600 MB/s, though the continuous transfer rates have to the researched).

However, here are the solutions I could come up with. By adequacy for my requirements – which included to be 128 GB or more in size:

  • SD card in SD card reader. This is the bes option on notebooks that come with a card reader, since it does not block any ExpressCard expansion port (reading your digital camera card is easily possible in the USB ports as well with SanDisk Ultra II SD Plus USB cards). However, it has to be tested still if the typical notebook card reader can utilize the speed of highest speed SD cards used here (else a ExpressCard/34 based card reader would be needed). Because for proper speed, we need the latest "U3" class SD cards. These are just being introduced around 2014-06, so prices will drop lateron. Currently available products:
  • SD card in ExpressCard/34 card reader. For data exchange, all readers can expose a SD card as a USB mass storage device, so no drivers on the host system are needed. The additional advantage of using a SD card is that you can also use them for other purposes (recording photos and HD video etc.). For proper speed, we need the U3 speed class cards again, see above.
  • Compact Flash card in ExpressCard/54 card reader. This is a good solution, since the speeds of Comact Flash cards are comparable to the ExpressCard SSDs while the Compact Flash format makes it simpler to find an adapter to USB mass storage (namely, a simple card reader). However, Compact Flash does not fit by width into ExpressCard/34, so your notebook will need the larger ExpressCard/54 slot [see]. Speeds of Compact Flash are great, and for data exchange you can carry a USB card reader that exposes the Compac Flash card as USB mass storage (means it needs no driver). But high-speed, large-capacity cards are also expensive; recommendable products:
    • A moderately priced one is the Hama CF 128 GB (up to 150 MB/s read, so maybe 40 MB/s continuous write?) for about 145 EUR.
  • ExpressCard/34 form factor SSD, plus USB adapter. This is the best option in terms of speed. It can be connected to other computers using a USB to ExpressCard adapter (overview; also see DeLock 61714 etc.). ExpressCard allows cards to use an USB 2.0, USB 3.0 or PCIe mode, and not all these adapters also support PCIe (the Digigear PCU10 does however, but needs manual configuration for PCIe mode). So check first which mode is used by the ExpressCard SSD – anything above 60 MB/s means USB 3.0 or PCIe mode. For the connectors etc., see the ExpressCard standards document.
    Available products 64 GB and larger:​

     

  • Small "knobby" USB flash drive. If your notebook has a USB 3.0 port or you can add one via an ExpressCard/34 slot or similar, the achievable speed of this solution can be good, too. It seems that there is still no 128 GB flash drive in "knob" size. A hack to get over this would be to combine two of them physically so that they fit exactly into two adjacent USB ports on your notebook (and with a small cable-type USB hub to other computers' USB ports). Available options:
  • microSD card in "knobby" microSD card reader. This option works even up to 128 GB (with 128 GB microSD cards being available). However, these cards tend to be slow. Available models: just one so far
  • Compact Flash card in 1.8" or 2.5" SATA drive bay. This is essentially using a Compact Flash card as a removable, rugged SSD. There are Compact Flash to SATA adapters for both 1.8" and 2.5" SATA SSD form factor (example: StarGat SAT2C). Also this solution will work only if your notebook can be modified in a way to allow drawing out the card from the HDD bay. And it's not really clear how the computer or operating system might react when extracting the Compact Flash card for example during ACPI Suspend mode. Also when using the system disk bay for this type of SSD, we cannot physically separate operating system and personal files as intended here.
  • miniPCIe SSD in ExpressCard and SATA external enclosure. This is more a theoretical option since you do not want to handle a bare PCB with your data on it just to connect it to a computer that does not have an ExpressCard slot. But it would work, as there are converters for miniPCIe to ExpressCard and miniPCIe to SATA.

If you are using Plesk as a server control panel and took part in all updates starting with some Plesk 10 version, you will notice that your recent version (Plesk 11.5 currently) has a total mess with respect to the web hosting directory layout. That's because during some of these updates, Plesk migrated its hosting directory structure, and now the old and new ones are mixed.

The new Plesk web hosting directory layout

Say you have two subscriptions (subscribers being the same or different customers, or even you as an admin). Each subscription has one subdomain and one additional domain. Then the web hosting directory layout would be like this (using placeholder domain names here):

  • /var/www/vhosts/subscr1-maindomain.com/subscr1-maindomain.com/
  • /var/www/vhosts/subscr1-maindomain.com/sub.subscr1-domain.com/
  • /var/www/vhosts/subscr1-maindomain.com/subscr1-otherdomain.com/
  • /var/www/vhosts/subscr2-maindomain.com/subscr2-maindomain.com/
  • /var/www/vhosts/subscr2-maindomain.com/sub.subscr2-domain/
  • /var/www/vhosts/subscr2-maindomain.com/subscr2-otherdomain.com/

Plus, in /var/www/vhosts/system/ there will be a directory for each domain and subdomain, named with its domain name, and containing subdirectories conf/, logs/ and statistics/.

Recommendation for using this directory layout

One subscription per (sub)domain. We want a directory structure that is logical and consistent, so it's not acceptable to find directories for some but not all domains inside /var/www/vhosts/. The only solution I could come up with so far is creating a new subscription for every single domain and subdomain that you want to host. And then to name the document root directory for hosting the main domain of that subscription always htdocs rather than with the domain name.

Problem: if you have a domain-restricted Plesk licence. With this method, you will run into problem depending on your Plesk licence. For example, if you have the typical "10 domains" licence. Because in this setup, every subdomain is created as its own Plesk domain record, so taking up one of the 10 available domains. While creating Plesk subdomain records for cases where their parent domain is also hosted in Plesk would not take up anything from your licence. But in my view: cleanliness first 🙂 And it's another reason to move to a recommendable, free and open source web hosting control panel.

Removing old, legacy elements. The following elements of the directory structure are probably legacy ones from the earlier Plesk versions and can be safely removed. However, I did not test this, so rename them first and check if everything still works before removing them:

  • /var/www/vhosts/example.com/subdomains/
  • /var/www/vhosts/example.com/conf/ (a symbolic link)
  • /var/www/vhosts/sub.example.com/ (in cases where this contains only conf/, logs/, statistics/ for a domain or subdomain that is an add-on domain for a subscription with another main domain, and has its document root inside that main domain's directory)
  • /var/www/vhosts/httpsdocs/ (probably not needed, as the new place seems to be /var/www/vhosts/default/httpsdocs/)

This is my list of Android applications that I recommend everyone to have and install on own phones and phones prepared for clients. Of course everybody will need more apps for specialized purposes, but these ones should be relevant for every user.

Selection criteria for this list:

  • only one selected best app for every purpose
  • any functional open source alternative is preferred to a polished proprietary software; proprietary licences are marked in red to indicate looking for a free & open replacement still
  • decentralized open communication protocols (e-mail, XMPP, SIP) are preferred to commercial centralized applications (Skype, Facebook)
  • communication apps should be privacy-enabled
  • the app selection should enable self organization and group organization, esp. for grassroots movements like social change activists

The "state" column contains my personal installation state info – just ignore 😉

state name FDroid Google Play description
notes
licence
x 3G/4G Speed Optimizer   Play Allows settings for faster 3G download and upload speeds, and for adjusting distribution between both. freeware
x 4EXT Recovery Control n/a Play For managing the 4EXT Recovery mode from inside an app. 4EXT Updater is a simpler, gratis alternative. commercial
x AdAway F-Droid   Removes all the pesky in-app advertisements. GPLv3
  Advanced Task Killer Free        
  ADW        
x Alarm Klock F-Droid   Full-featured alarm clock to help you wake up. Apache2
  aLogcat        
  AnaCam        
  AndroSS        
  AndStatus        
x Angulo F-Droid   For measuring angles and distances, also useable as a water-level. GPLv3
  AnyMemo        
  AnySoftKeyboard        
x APV PDF Viewer F-Droid     GPLv3
x Arity F-Droid   Scientific calculator. Apache2
x Barcode Scanner F-Droid   Complete barcode and QR code scanning app, employed by other apps as well. Apache2
  BLUE LINE        
n/y BlueGps F-Droid n/a Allows to connect an external GPS device and make it usable in e.g. OsmAnd~ via the "mock location provider" option. Means faster, more precise and higher coverage GPS, longer battery runtime, and allows placing the GPS receiver outside the car. To be evaluated. GPLv3
x Bluetooth GPS n/a Play Same as BlueGps, but non-free. Good quality though. freeware
  Busybox   Play Installer for Busybox, a boiled down version of GNU Core Utils. Needed for many other root apps, incl. 4EXT Recovery Updater. GPL 2
  CamPainter Lite        
x ChatSecure F-Droid   Encrypted XMPP chat client, including off-the-record features. Apache2
x Clock n/a n/a This app (com.android.deskclock) comes pre-installed with Android and is well done, including clock, alarm clock, countdown clock and stop watch. Only the alarm clock misses some features, so we also have Alarm Klock. open source
  ColorDict        
  Compass        
  ConnectBot        
  Cool Reader        
  Diktofon        
x Evernote   Play Note taking and organizing system that syncs to desktop and web. freeware
  Every Locale        
x F-Droid F-Droid   Package manager / marketplace for all the Android open source software recommended here. GPLv3+
  Fake Dawn        
  FAST App Search Tool        
x Faster GPS F-Droid   Enables getting faster GPS fixes by selecting a NTP server near you. GPLv3
x Firefox F-Droid   Web browser, mobile edition. MPL2
  FolderSync        
  Freemind Viewer        
x GetBack GPS F-Droid   Finding a way back to a "home location" with GPS guidance. Hopefully better than Point, which can be inaccurate. GPLv3+
  Google Play services        
x Google Play Store        
x K-9 Mail F-Droid   Full-featured e-mail client. Apache2
  KeepTrack        
  Ki Freemind        
  Maniana        
x Mathdroid F-Droid   Handy calculator with history. GPLv3
  MobileOrg        
  MobileWebCam        
x MrWhite F-Droid      
x Network Discovery F-Droid   A network utility: discover hosts inside a network you are connected to, and open ports on these hosts. GPLv2
  Notification Plus        
  Obsqr        
  Offline Calendar        
  OS Monitor no-gmaps        
x Osmand~ F-Droid   Super great offline map viewing and routing application. Uses OpenStreetMap data. GPLv3
x OSMTracker F-Droid   GPX tracker that allows note-taking, esp. also meant as a help for OpenStreetMap mapping work. GPLv3
x Pedometer F-Droid     GPLv3
  Blockchain        
  QuickDic        
  Remuco        
  Ruler        
x SatStat F-Droid   GPS, sensor and network status information. So for most purposes, a more low-level graphing app like Sensor Readout is not needed. GPLv3+
  Sca Lite        
  Search based launcher        
  Search Light F-Droid   Flashlight using the camera LEDs. Screen is on at the same time, so MrWhite is also installed as a less energy hungry alternative.  
  Speech Trainer        
  Super Ruler Free        
  TallyBee        
  Tallyphant        
  Terminal Emulator        
  Theme Chooser Themes        
  Timesheet        
  Units        
  Vanilla Music        
  Who Has My Stuff?        
  WiGLE Wifi Wardriving        
  WikiDroyd        
  Wikipedia        
x Word Lens Translator n/a Play Travel app that translates words recognized in the live camera view between five languages and English. All happens offline, no network connection is used for anything. commercial
n/y WordPress F-Droid     GPL
  YAAB        
  YouTube        

I am still looking for apps for:

  • SMS backup application (there is SMS Backup+, but it cannot save to SD card)
  • accessing microblogging services incl. Identi.ca and Twitter
  • App for power saving while travelling: switch mobile network off (airplane mode) while the phone moves by car or train, as determined by GPS from speed and location. And only check every 30 min for missed calls.
  • Find a powerful, high usability text editor for Android. Ideally with hierarchical mode that is compatible with Freemind's double-indent text exports.
  • Find or develop an app that unifies all possible action on clipboard content, can be called from every app and stores clipboard history, and analyzes clipboard content and history to suggest relevant actions (like "send wire payment to").

Usually, you can have mentions with e-mail notifications using the Drupal mentions module, or mentions with auto-suggest using the Drupal ckeditor_mentions module, but not both. But it turns out you can, indeed, by combining both modules (see also). Here are detailed instructions how to set thus up.

Recommended: Using Message Notify for message sending

  1. Install and enable required modules: rules, mentions, ckeditor_mentions, message_notify.
  2. Configure the mentions module (in /admin/config/content/mentions) to  use "Input: Prefix: @" and "Input: Suffix: " (nothing). This corresponds to how usernames are displayed after letting ckeditor_mentions insert a suggestion.
  3. Make sure your usernames are set up to not contain spaces, because the @username format does not allow usernames with spaces to be recognized.
  4. Enable permission "Bypass Rules access control" for your user's user group, or you will get a "Access violation! You have insufficient access permissions to edit this configuration." error when trying to create the "mention:entity of type …" Rule condition below.
  5. In the admin menu, go to "Structure -> Message types" (/admin/structure/messages) and create a new message type. It is simplest to clone it from one that is meant for use in e-mails already, such as any Commons Notify message type (if you are a Drupal Commons user).
  6. Edit your new message type to your needs. You can use the following replacement tokens in both subject and body fields:
    1. [message:user:name] Username of the user to whom the message is sent.
    2. [mention:author:name] Username of the user who created the mention.
    3. [mention:entity:…]
  7. Create a new Rule (say, named message_on_mention_in_node) like this:
    1. Specify the rule's trigger to be: "React on event: After a new mention is created".
    2. Add a condition specifying "mention:entity is of type: Node".
    3. Add an action "Create a new entity". [TODO]
    4. Add an action "Provide a data value". [TODO]
    5. Add an action "Save entity". [TODO]
    6. Add an action "Send message with Message Notify" [TODO]
    7. The four actions in your rule should now look similar to this screenshot.
  8. Create a new Rule messgae_on_mention_in_comment, in analogy to the above but with the condition using entity type "comment".

Alternative: Using PHP code for message sending

It is also possible to create a rule that will send the required e-mail using custom PHP code. However, since this requires enabling the "PHP Filter (php)" module, it is not considered good practice security-wise. However, if you want to use this route, you can do so as follows:

  1. Follow steps 1-3 from the above list.
  2. Create a Rule email_on_mention_with_php.
    1. Specify the rule's trigger to be: "React on event: After a new mention is created".
    2. Add an action "Send mail", set its "To:" field to "mention:user:mail" and the Subject field to something similar to the following text:

Hello [mention:user:field-name-first] [mention:user:field-name-last],

you got mentioned by [mention:author:field-name-first] [mention:author:field-name-last] here:

<?php
if ($mention->entity_type == 'comment') {
  echo '"' . $mention->entity->subject . '"' . "\n";
  echo $mention->entity->comment_body[‘und’][0][‘safe_value’];
}
elseif ($mention->entity_type == 'node') {
  echo '"' . $mention->entity->title . '"' . "\n";
  echo $mention->entity->body[‘und’][0][‘safe_value’];
}
else {
  echo '<no preview available>';
}
?>

You can view the mention and comment on it here:
<?php
if ($mention->entity_type == 'comment') {
  $cid = $mention->entity->cid;
  echo("http://edgeryders.eu/comment/$cid#comment-$cid");
}
elseif ($mention->entity_type == 'node') {
  $nid = $mention->entity->nid;
  echo("http://edgeryders.eu/node/$nid");
}
else {
  echo '<no URL available>';
}
?>

best regards,
xxx

<?php
// echo("——————-\n" . 'Debugging information: $mention = ');
// print_r($mention);
?>

 

Forwarding web traffic

This can come in very helpful when switching servers and wanting to avoid any downtime while the updated DNS entries propagate over 48 hours through the Internet. This is esp. important for websites with

  1. Make sure you meet the requirements. You need:
    1. Your old server has to be a Linux / Unix based server, so that rinetd can be installed.
    2. root access to your old server for installing rinetd
  2. Copy over your websites and databases. You might want to put community websites into maintenance mode first because it's impractical and spammy to write hundreds of users an e-mail saying they should not add content over the next half an hour …
  3. Install rinetd. On your old server, execute (assuming Debian or Ubuntu): apt-get install rinetd
  4. Configure rinetd. For web traffic, the following lines in /etc/rinetd.conf are enough (see the docs for details, and use your new server's IP address instead of xxx.xxx.xxx.xxx of course):
    0.0.0.0  80  xxx.xxx.xxx.xxx  80
    0.0.0.0 443  xxx.xxx.xxx.xxx 443
  5. Stop your webserver. So, do something like service apache2 stop. rinetd can only work successfully if nothing else is connected to the ports it shoud bind to. If there is something else listening, it will result in no redirections and a huge rinetd log file containing messages like 22/Mar/2013:19:35:47 0.0.0.0 0 (null) 0 0 0 accept-failed – [source].
  6. Start rinetd. That's simply service rinetd restart.
  7. Test it. If your websites still work, they are now served by the new server, since you disabled the webserver on the old one.
  8. Change your DNS records. To make everyones requests switch to the new server, point your sites DNS records to the new server.

Forwarding e-mail traffic

You can now migrate e-mails with a similar approach, something along these lines:

netstat -t -l -p
service courier-imapd stop
service courier-imaps stop
service courier-pop3d stop
service courier-pop3s stop
service postfix stop

Then add more forwarding rules in /etc/rinetd.conf – in my case I now have these in total:

# any private mail system
0.0.0.0         24        xxx.xxx.xxx.xxx   24
# SMTP
0.0.0.0         25        xxx.xxx.xxx.xxx   25
# Secure SMTP
0.0.0.0         26        xxx.xxx.xxx.xxx   26
# HTTP
0.0.0.0         80        xxx.xxx.xxx.xxx   80
# POP3
0.0.0.0        110        xxx.xxx.xxx.xxx  110
# IMAP
0.0.0.0        143        xxx.xxx.xxx.xxx  143
# HTTPS
0.0.0.0        443        xxx.xxx.xxx.xxx  443
# IMAP over TLS
0.0.0.0        993        xxx.xxx.xxx.xxx  993
# POP3 over TLS
0.0.0.0        995        xxx.xxx.xxx.xxx  995

And finally restart rinetd to make it pick up the new config:

service rinetd restart

Alternative approaches

  • Using iptables. It also is about redirecting traffic at NAT / TCP level. However, this it is more difficult to understand and to configure, since it is meant as a full-featured software firewall.
  • Using remote databases. If all your sites support using a database on a different host (Drupal, WordPress etc. all can do that), you can do it as follows. The advantage is, you can move your sites one by one, so can take time to iron out any issues on the new server without having to prepare that for all sites at once (which would mean copying over databases twice, once to test and once to take in new changes). A disadvantage of this is that it only works for websites, not for e-mails. Step by step:
    1. Put your site in maintenance mode for a short time.
    2. Move your site and its database over to the new server.
    3. Test your site on the new server by bending DNS at your own machine (edit /etc/hosts if you are on Linux).
    4. Edit your site configuration on your old server to use the remote databases on the new one.
    5. Optionally (if your site has user-generated files), set up an rsync / scp / ftp synchronization job to push or pull new and updated files to the new server, called by a cron job every 15 minutes or so.
    6. Disable maintenance mode again and test if the site on the old server still works (remember to disable DNS bending again!).
    7. Adapt your DNS configuration to point  to the new server. After 48 hours at latest, nobody will access the old server any more.

 

You can follow all the steps of the official HowTo “The Perfect Server – Ubuntu 13.04 (Apache2, BIND, Dovecot, ISPConfig 3)“. To get it to work afterwards on Ubuntu 13.10, use these additional steps:

  1. Switch to a .vhost naming scheme for Apache2 site configurations. In Apache2 2.4, only the sites-enabled/*.conf files will be loaded automatically, missing out ISPConfig’s /etc/apache2/sites-enabled/*.vhost. Usually we would rename them to .conf, but because .vhost files are also generated automatically for every website created in ISPConfig, we instead adapt what Apache2 loads. Because that is less work than adapting the source code of ISPConfig. Note in the following commands, both for sites and further down for config snippets, that it’s important to keep the original apps.vhost and ispconfig.vhost included at the start of the site inclusion process, so let’s keep the “000-” and “100-” prefixes used by ISPConfig [source].
    1. In /etc/apache2/apache2.conf, replace line “IncludeOptional sites-enabled/*.conf” with “IncludeOptional sites-enabled/*.vhost“.
    2. mv /etc/apache2/sites-available/000-default.conf /etc/apache2/sites-available/000-default.vhost;
    3. mv /etc/apache2/sites-enabled/000-default.conf /etc/apache2/sites-enabled/000-default.vhost;
    4. rm /etc/apache2/sites-enabled/000-apps.vhost;
    5. mv /etc/apache2/sites-available/apps.vhost /etc/apache2/sites-available/000-ispconfig-apps.vhost;
    6. ln -s ../sites-available/000-ispconfig-apps.vhost /etc/apache2/sites-enabled/000-ispconfig-apps.vhost;
    7. rm /etc/apache2/sites-enabled/000-ispconfig.vhost;
    8. mv /etc/apache2/sites-available/ispconfig.vhost /etc/apache2/sites-available/000-ispconfig.vhost;
    9. ln -s ../sites-available/000-ispconfig.vhost /etc/apache2/sites-enabled/000-ispconfig.vhost;
  2. Move and rename the ISPConfig Apache2 config file. In addition, the ISPConfig .conf files which are also present actually contain configuration information and not vhost directives for sites, so a2ensite and a2dissite cannot handle them. They are now expected in /etc/apache2/conf-available, to be handled by a2enconf and a2disconf. So let’s rearrange things a bit to solve this (note that we use a “.local” suffix to indicate local configuration additions, not provided via OS packages):
    1. rm /etc/apache2/sites-enabled/000-ispconfig.conf;
    2. mv /etc/apache2/sites-available/ispconfig.conf /etc/apache2/conf-available/000-ispconfig.local.conf;
    3. a2enconf 000-ispconfig.local;
  3. (optional) Fix a2ensite and a2dissite. Due to our transition to the “.vhost” naming scheme for Apache2 site configurations, a2ensite / a2dissite do not work any more because they expect a .vhost suffix. Not too bad because one will enable and disable most sites within ISPConfig anyway. However, we can also repair this by supplying a local alternative with a fix in the source code:
    1. cp /usr/sbin/a2ensite /usr/local/bin;
    2. ln -s a2ensite /usr/local/bin/a2dissite;
    3. In file /usr/local/bin/a2ensite, replace line 60 “$sffx   = ‘.conf’;” with “$sffx   = ‘.vhost’;“.
    4. Ensure with which a2ensite and which a2dissite that your local alternatives are in use now.
    5. Unfortunately, this fix only works for sites not handled by ISPConfig. To also handle those created by ISPConfig, one has to find a solution for the fact that ISPConfig creates the symlinks with a “100-” prefix while the site configurations in sites-available/ miss this. The simplest is to provide an alternative with that prefix in sites-available/ as well, using something like:
      ln -s example.com.vhost /etc/apache2/sites-available/100-example.com.vhost;
  4. Fix a syntax error about directory options. Apache would now complain about a syntax error: “Either all Options must start with + or -, or no Option may.” So add a “+” character before all options missing a “+” or “-” prefix in:
    /etc/apache2/sites-available/000-ispconfig-apps.conf line 30
    /etc/apache2/sites-available/000-ispconfig.conf line 20
  5. Remove the NameVirtualHost directives. They don’t really hurt, but are effectless and deprecated in Apache 2.4, generating warnings when restarting Apache. So remove them from:
    /etc/apache2/sites-available/000-ispconfig.conf line 8
    /etc/apache2/conf-available/000-ispconfig.conf lines 53 and 54
  6. Reload the Apache configuration.
    service apache2 reload;
  7. Test ISPConfig and configure it. If you have enabled SSL for the ISPConfig site during the installation configure script, try visiting your server’s FQDN at port 8080 in your browser (like browsing to https://hostname.example.com:8080/). Initial username and password are “admin” and “admin” – remember to change them immediately. The remaining configuration happens from inside the ISPConfig 3 panel.

We needed a fast, dedicated server to solve optimization problems. So let’s install COIN-OR CBC (the MLP optimization solver we use). We start from a stock Ubuntu 13.10 machine, as it will happen to you when deploying on Amazon EC2 Spot Instances, like we do.

The easiest way to get COIN-OR OS and CBC installed on a common Linux platform is to install the “full release” binary packages, which contain the binaries from all major COIN-OR projects. Just download the latest COIN-OR-*.tgz package from the CoinAll repo, unpack it, and link the files in its bin/ subdirectory into your path with something like ln -s bin/* /usr/local/bin/.

However, we want to install from source here, which gives us some advantages (mostly following reasoning found here):

  • The most important reason is that COIN-OR cannot distribute GPL-licenced source or binaries with its code. It includes download scripts for these however. So to get goodies like ASL (the AMPL input format library we desperately need for CBC here), GMPL, command completion, command history, Haskell libraries etc. one has to install from source.
  • The installation uses the Debian package management system for simple modifications, updates and removal. (Some COIN-OR packages are available in the Ubuntu archives, but not all from the COIN-OR distribution – not OS for example, which we need here. Also these packages tend to be a bit outdated.)
  • We can optimize COIN-OR to our needs with the many compilation options of COIN-OR. For example, enabling multi-threading for the CBC solver, or optimizatios for specific CPUs / hardware.
  • Bug fixes can be incorporated immediately instead of waiting for a new COIN-OR release.

How to install all major COIN-OR packages from source, according to the CoinAll distribution build notes and the guidlines from the T. K. Ralphs presentation:

  1. Install some needed developer tools. Note that csh is needed for DyLP/doc/build_dylpdoc, which is called in make distclean for example.
    sudo apt-get install patch make pkg-config wget subversion checkinstall csh gcc g++ gfortran
  2. Install some libraries needed by some COIN-OR packages. (matplotlib is needed for the GiMPy tests to complete.)
    sudo apt-get install python-matplotlib
  3. Download the latest stable version. This probably will get you the latest maintenance release, so probably currently 1.7.4 as seen in the CoinAll list of binaries.
    svn co http://projects.coin-or.org/svn/CoinBinary/CoinAll/stable/1.7 CoinAll-1.7
  4. Add necessary and desired third-party source code. Some code could not be included into the COIN-OR distribution because of licence incompatibility (for example with the GPL). But COIN-OR provides nice download scripts to add these components:
    1. Add the MUMPS public domain linear solver. If present, it will be used in the compilation of Ipopt. If not present, Ipopt will be compiled without solvers, which will cause make test fail for Ipopt with a message like “selected linear solver ma27 not available” [source].
      cd ~/CoinAll-1.7/ThirdParty/Mumps
      ./get.Mumps
    2. Add Glpk. This is needed by CBC in order to read GMPL input – that’s exactly what we need, since it is the only way to use an algebraic modeling language for comfortable, high-level input to CBC. GMPL (also called MathProg) is actually a subset of the widespread AMPL [source], so many models in AMPL should also be expressable with it.
      cd ~/CoinAll-1.7/ThirdParty/Glpk
      ./get.Glpk
    3. Add the ASL library. Without this being present, CBC would be built without AMPL .nl input format support – trying to call it with a .nl file would then result in an error “nl file specified locally but ASL not present”. But we want to have the nice simple AMPL input format. So to get it included, we do according to the CBC FAQ:
      cd ~/CoinAll-1.7/ThirdParty/ASL
      ./get.ASL
    4. Add BLAS. Adding Blas (andLapack, see below) to CBC gives better numerical stability to CBC’s LP solver [source, p.5] – whatever that means. Blas is also needed for compiling one of the other COIN-OR projects; it can be added with apt-get install libblas3 libblas-doc libblas-dev but let’s keep here with the mechanism supplied by COIN-OR to get the latest version:
      cd ~/CoinAll-1.7/ThirdParty/Blas
      ./get.Blas
    5. Add Lapack. Second component in addition to Blas for improving the numerical stability of CBC. To add it:
      cd ~/CoinAll-1.7/ThirdParty/Lapack
      ./get.Lapack
    6. (optional) Add METIS. METIS is a library for graph partitioning, and I don’t know if CBC can profit from it. But since it’s the last of the ThirdParty packages that are free to use for all, let’s go ahead:
      cd ~/CoinAll-1.7/ThirdParty/Metis
      ./get.Metis
  5. Configure it all. In the configure command below, the LDFLAGS=-Wl,--no-as-needed; is to solve problems caused probably by some library symbols now being versioned in current versions of Ubuntu. These problems show up with error messages saying “undefined reference to dlopen, dlclose, dlerror, dlsym”. The LDFLAGS setting is a solution from the ipopt mailing list, and while it works for me I don’t quite understand it. It is only needed for ./configure in the sub-projects OS, DyLP and IPopt. But it does not hurt to apply it globally, so for simplicity we do just that. Also, we configure CBC to support parallel execution. This is done by adding the --enable-cbc-parallel compile-time option [source]. Again, it does not hurt to add it to the global ./configure, it will only affect CBC (and should result in a line “Cbc multithreading enabled” in build/Cbc/configure.log).
    mkdir ~/CoinAll-1.7/build;
    cd ~/CoinAll-1.7/build/;
    ../configure --enable-gnu-packages --enable-cbc-parallel -C --prefix=/usr/local/share/coin-or LDFLAGS=-Wl,--no-as-needed;
  6. Compile it all.
    cd ~/CoinAll-1.7/build/;
    make -j 2;
  7. Adapt the tests for a headless server. Some tests will fail without an X Server. We run this on a headless server, so we have to modify the tests (or use Xvfb, but that is a hack as well …).
    1. In ~/CoinAll-1.7/GiMPy/test/test_algorithms.py, remove line produce_graphs() (actually the very last one).
    2. Solve a similar problem with ~/CoinAll-1.7/GrUMPy/test/test_bb.py.
    3. TODO There may be more issues like this. I did not finish trying to get all tests to run.
  8. Run the tests. (Might still fail at some point when used on a headless server, as per the last point. But you will get through with most at least.)
    make test;
  9. Create a directory for the installation.
    sudo mkdir /usr/local/share/coin-or
  10. Create a Debian package and install it. Installing is included in the following command if all works well. In total, it can take a while, 20 minutes even, depending on your machine. I named mine coinor-coinall.
    sudo checkinstall make install;
  11. Put the binaries into the path. Because this is not done by the installation. Let’s use symlinks:
    cd /usr/local/bin && sudo ln -s ../share/coin-or/bin/* .

TODO The packages built this way can be installed on any Ubuntu 13.10, 14.04 or perhaps newer version, but they miss to list some dependencies, so these should be added manually in the “checkinstall make install” step. Until then, make sure to provide the packages yourself with:

sudo apt-get install gfortran