Category Archives: debian

How I learnt about linux’ “OOM Killer”

I have a low-end VPS (Virtual Private Server), or at least: it used to be low-end, now it’s at least one step above lowest and cheapest.

On this server I’m running various personal stuff: email (with spamassassin), some public web pages, and some private web pages with various small webapps for the household (e.g. the “weekly allowance app” registering work done and payments made).

I had noticed occasional slow startups of the webapps, and in particular in September this year, when I was demonstrating/showing off the webapps at this year’s JavaZone the demos were less than impressive since the webapps took ages to load.

I was quick to blame the slowness on the wi-fi, but as it turns out, that may have been completely unfair.

The webapps had no performance issues on my development machine even when running with a full desktop, IDE and other stuff.  The webapps on the VPS also seemed to have no performance issues once they loaded.

I thought “this is something I definitely will have to look into at some later time…” and then moved on with doing more interesting stuff, i.e. basically anything other than figuring out why the webapps on a VPS were slow to start.

But then the webapps started failing nightly and I had to look into the problem.

What I saw in the logs was that the reason the webapps were broken in the morning was that they were stuck waitning for a liquibase database changelog lock that never was released.

Liquibase is how my webapps set up and update database schemas. Every time a webapp starts it connects to the database and checks what liquibase scripts that have been run against that database and applies the ones that have not already been run. The list of scripts that have been run is a tabled called databasechangelog. And to avoid having more than one liquibase client attempting to modify the database schema, liquibase uses a different table called databasechangeloglock to moderate write access to the database,

I.e. the databasechangeloglock is just a database table that has one or 0 rows. A liquibase client tries to insert a lock into the table at startup and waits and retries if this fails (and eventually completely fails).

In my case the webapps were failing because they were hanging at startup, trying to get a liquibase lock and failing to get one and were hanging in limbo and never completing their startup process. Manually clearing the lock from the table and restarting the webapps made the webapps start up normally. However, the next day the webapps were failing again for the same reason: the webapps were stuck waiting for a liquibase lock.

I initially suspected errors in my code, specifically in the liquibase setup. But googling for similar problems, code examination and debugging revealed nothing. I found nothing because there was nothing to be found.  The actual cause of the problem had nothing to do with the code or with liquibase.

I run my webapps in an instance of apache karaf that is started and controlled by systemd. And I saw that karaf was restarted 06:30 (or close to 06:30) every morning. So my next theory was that systemd for some reason decided to restart karaf 06:30 every morning.

No google searches for similar symptoms found anything interesting.

So I presented my problem to a mailing list with network and linux/unix experts and got back two questions:

  1. Was something else started at the same time?
  2. Did that something else use a lot of memory and trigger the OOM killer?

And that turned out to be the case.

I have been using and maintaining UNIX systems since the mid to late 80ies and setting up and using and maintaining linux systems since the mid to late 90ies, but this was the first time I’d heard of the OOM killer.

The OOM killer has been around for a while (the oldest mention I’ve found is from 2009), but I’ve never encountered it before.

The reason I’ve never encountered it before is that I’ve mostly dealt with physical machines. Back in the 80ies I was told that having approximately two and a half times physical memory was a good rule of thumb for scaling swap space, so that’s a rule I’ve followed ever since (keeping the ratio as the number of megabytes increased, eventually turning into gigabytes).

And when you have two and a half times the physical memory as a fallback, you never encounter the conditions that make the OOM killer come alive.  Everything slows down and the computer starts trashing before the condtions that triggers the OOM killer comes into play.

The VPS on the other hand, has no swap space. And with the original minimum configuration (1 CPU core, 1GB of memory), if it had been a boat it would have been said to be riding low in the water. It was constantly running at a little less than the available 1GB. And if nothing special happened, everything ran just fine.

But when something extraordinary happened, such as e.g. spamassassin’s spamd starting at 06:30 and requiring more memory than was available, then OOM started looking for a juicy fat process to kill, and the apache karaf process was a prime cadidate (perhaps because of “apache” in its name combined with OOM killer’s notorious hatred of feathers?).

And then systemd discovered that one of it’s services had died and immediately tried to restart it, only to have OOM killer shoot it down, and this continued for quite a while.

And in one of the attempted restarts, the webapp got far enough to set the databasechangeloglock before it was rudely shot down, and the next time(s) it was attempted started it got stuck waiting for a lock that would never be released.

The solution was to bump the memory to the next step, i.e. from 1GB to 2GB. Most of the time the VPS is running at the same load as before (i.e. slightly below 1GB) but now a process that suddenly requires a lot of memory no longer triggers the OOM killer and everything’s fine.  Also the available memory is used for buff/cache and everything becomes much faster.

I bumped the memory 8 weeks ago and the problem hasn’t occurred again, so it looks like (so far) the problem has been solved.

Faking a debian repository for package development

I use aptly to deliver my unofficial debian packages both to myself and others that might be interested.

However I’ve found that using aptly to do package development is a bad idea, because you can’t (by design, probably) overwrite packages in an aptly archive.  You can only create new versions.

For some installation tests it’s OK to use “dpkg –install”. But if your package needs to pull in depdencies, or if you wish to test a package upgrade, you need to use APT.

This article explains how to create a fake debian repository for use in package development.

Initial setup

Initial setup:

  1. Create the repository directory (this should be done as the same user you use to build the package)
    mkdir -p /tmp/repo
  2. Open /etc/apt/sources.list in a text editor and add the following (you need to be root to do this)
    # Test install apt archive
    deb file:///tmp repo/
    

Add new package to the repo

This the development cycle:

  1. Build the package (the example builds karaf)
    cd ~/git/karaf-debian/
    dpkg-buildpackage
  2. Copy the package to the fake repo (this example uses karaf, replace with your own package and package version):
    cp ~/git/karaf_4.1.5-0~9.30_all.deb /tmp/repo
  3. Generate a Packages file
    (cd /tmp; dpkg-scanpackages repo >repo/Packages)

Upgrading an existing package

This has to be done as root:

  1. First update APTs database of packages and archives
    apt-get update
  2. Run an upgrade with the following command
    apt-get dist-upgrade
  3. There will will be a question for if you wish to continue, to continue is the default, just press ENTER
    Do you want to continue? [Y/n]
  4. There will be a warning about that the package to be installed cannot be authenticated, the default here is not to install, so press “y” (without the quotes) to continue
    Install these packages without verification? [y/N]

Installing a package

This is used to e.g. test that a package is able to install its dependencies.

These operations have to be done as root:

  1. First update APT’s database of packages and archives
    apt-get update
  2. Use the following command to install a package and its dependencies, this example installs apache karaf:
    apt-get install karaf
  3. If the package pulls in new dependencies there will be a prompt for if you wish to continue. The default is to continue, so just press ENTER
    Do you want to continue? [Y/n]
  4. There will be a warning about that the package to be installed cannot be authenticated, the default here is not to install, so press “y” (without the quotes) to continue
    Install these packages without verification? [y/N]

Installing apache karaf on debian

Until the RFP (Request For Packaging) bug for karaf in the debian bug tracker is resolved, here is an APT archive with a karaf package for debian (architecture “all”).  The package is created using native debian packaging tools, and built from a source tarball and the APT archive itself is created, using aptly.

The package has been tested on Debian 9 “stretch” (the current stable), amd64.

Do the following commands as root on a debian GNU/linux system:

  1. Add the key for the APT archive
    wget -O - https://apt.bang.priv.no/apt_pub.gpg | apt-key add -
  2. Open the /etc/apt/sources.list file in a text editor, and add the following lines:
    # APT archive for apache karaf
    deb http://apt.bang.priv.no/public stable main
  3. Install karaf with apt-get
    apt-get update
    apt-get install openjdk-8-jdk karaf
  4. Log in with SSH (password is “karaf” (without the quotes)) and try giving some commands:
    ssh -p 8101 karaf@localhost

 

Packaging karaf with native debian packaging tools

Note! This is an improvement over the packaging in  Installing apache karaf on debian stretch, this package is packaged using native debian packaging tools instead of fpm, and is built from the karaf source tarball instead of the karaf binary tarball.

Apache karaf is an OSGi container and application server that is provisioned from maven, and has an ssh server. Basically it is possible to start an empty karaf, ssh in and give some commands to install an application using maven.

There still isn’t a native .deb package on maven (see  the RFP (Request For Packaging) bug for karaf in the debian bug tracker), but this package can be installed from my own maven repository.

The packacing projecct can be found on github: https://github.com/steinarb/karaf-debian

Procedure to build the package

  1. Install the required build tools
    apt-get update
    apt-get install openjdk-sdk git maven-debian-helper devscripts
  2. Clone the karaf package repository
    mkdir -p ~/git
    cd ~/git/
    git clone https://github.com/steinarb/karaf-debian.git
  3. Build the package
    cd ~/git/karaf-debian/
    dpkg-buildpackage

After this, there will be a karaf-*.deb package in the directory above the karaf-debian directory.

Setting up a debian package archive with aptly

This article describes how to set up a debian archive with aptly on a debian 9 “stretch” computer, served by an nginx web server.

Initial setup

  1. Add a DNS alias for your virtual nginx web site (outside of the scope of this blog post). The examples below assume that apt.mydomain.com is the DNS alias
  2. Install the required software, logged in as root, give the following command
    apt-get install gnupg pinentry-curses nginx aptly
  3. Logged in as your regular user, do the following:
    1. Create a gpg keyNote! It is a good idea to do the key generation when logged into a debian desktop and move the mouse about during generation, to get good random values for the key generation.Giving the following command at the command line
      gpg --full-generate-key
      1. At the prompt for key type, just press ENTER to select the default (RSA and RSA)
      2. At the prompt for key size, type “4096” (without the quotes) and press ENTER
      3. At the prompt for how long the key should be valid, type “0” without the quotes and press ENTER
      4. At the prompt for “Real name”, type your real name and press ENTER
      5. At the prompt for “Email address”, type your email address and press ENTER
      6. At the prompt for “Comment”, type the host name of your archive web server, e.g. “apt.mydomain.com” and press ENTER
      7. At the prompt “Change (N)ame, (C)omment, (E)mail or (O)kay/(Q)uit?”, type “O” (without the quotes) and press ENTER
      8. At the prompt for a passphrase, give a passphrase you will remember. You will be asked for this passphrase every time every time the repository is published
    2. Export the public key of the gpg key pair in a form that the web server can use
      1. First list the keys to find the id of the key
        steinar@lorenzo:~$ gpg --list-keys
        /home/steinar/.gnupg/pubring.gpg
        ---------------------------
        pub   rsa4096 2017-12-27 [SC]
              6B7638490707CCE365DF5415D2BA778731DC1EAC
        uid           [ultimate] Steinar Bang (apt.mydomain.com) <sb@dod.no>
        sub   rsa4096 2017-12-27 [E]
        
      2. Then use the id of the key to export the public key of the gpg key pair in a form that the web server can return
        gpg --output apt_pub.gpg --armor --export 6B7638490707CCE365DF5415D2BA778731DC1EAC
      3. Publish the key with the default GPG keyserver
        gpg --send-keys 6B7638490707CCE365DF5415D2BA778731DC1EAC
    3. Create a local repository “stable”
      aptly repo create -distribution="stable" stable
    4. Configure an architecture in the archive: open the ~/.aptly.conf file in a text editor, and change the line
      "architectures": [],

      to

      "architectures": ["amd64"],

      Note! Without a concrete architecture in place, aptly refuses to publish. So add an architecture here, even you are going to publish packages with architecture “all” (e.g. java, python, shell script). In the example I’m using “amd64” which, despite its name, is appropriate for modern 64 bit intel chips (i5 or i7 of various generations).

    5. Import a package into “stable” (the example uses the package built in Installing apache karaf on debian stretch)
      aptly repo add stable git/karaf-deb-packaging/karaf_4.1.4-1_all.deb
    6. Publish the archive (switch the gpg-key with the id of your actual repository key):
      aptly publish repo --gpg-key="6B7638490707CCE365DF5415D2BA778731DC1EAC" stable

      Note! If you get a time out error instead of a prompt for the GPG key passphrase, and you’re logged in remotely to the server, the reason could be that gpg tries to open a GUI pinentry tool. Switch to explictly using a curses-based pinentry and try the “aptly publish” command again. Do the command:

      update-alternatives --config pinentry

      and select “pinentry-curses” in “Manual mode”

  4. Log in as root and do the following
    1. Create a root directory for a new web server and copy in the public key used to sign the published achive
      mkdir -p /var/www-apt
      cp /home/steinar/apt_pub.gpg /var/www-apt/
    2. In a text editor, create the file /etc/nginx/sites-available/apt with the following content
      server {
      	listen 80;
      	listen [::]:80;
      
      	server_name apt.mydomain.com;
      	root /var/www-apt;
      	allow all;
      	autoindex on;
      
      	# Full access for everybody for the stable debian repo
      	location /public {
      		root /home/steinar/.aptly;
      		allow all;
      	}
      
      	# Allow access to the top level to be able to download the GPG key
      	location / {
      		allow all;
      	}
      }
      

      Note! I actually started out with also serving HTTPS and signing with letsencrypt, but as it turns out APT doesn’t support HTTPS out of the box, so there was no point in including it in this HOWTO

    3. Enable the site by creating a symlink and restarting nginx
      cd /etc/nginx/sites-enabled
      ln -s /etc/nginx/sites-available/apt .
      systemctl restart nginx

Your APT artchive is now up and running.

Use the new APT archive

To start using the APT archive, do the following on a debian computer:

  1. Log in as root
  2. Add the archive key
    wget -O - https://apt.mydomain.com/apt_pub.gpg | apt-key add -
  3. Add the archive by adding the following lines to /etc/apt/sources.list
    # My own apt archive
    deb http://apt.mydomain.com/public stable main
  4. Update APT to get the information from the new archive
    apt-get update
  5. Install a package from the archive
    apt-get install karaf

Future additions and updates of existing packages can be done as your regular user, with no need to log in as root during the process.

Publish a new version of a package

To update an existing package:

  1. Build a new version of the package
  2. Add the new version of package to the package archive
    aptly repo add stable git/karaf-deb-packaging/karaf_4.1.5-1_all.deb
  3. Update the publish the package archive (i.e. the package archive served by nginx)
    aptly publish update --gpg-key="6B7638490707CCE365DF5415D2BA778731DC1EAC" stable
  4. Do “apt-get update” on the computers using the archive
  5. Do “apt-get dist-upgrade” on the computers using the archive, and the package should be upgraded

Remove old versions of a package

To delete old versions of a package:

  1. Do a query to verify that the expression matches only the packages you want to delete
    aptly repo search stable 'karaf (<=4.1.4-4)'
  2. Remove all packages matching the query
    aptly repo remove stable 'karaf (<=4.1.4-4)'
  3. Remove old versions of the database (this is where the disk usage of the repository is reduced)
    aptly db cleanup

 

Installing apache karaf on debian stretch

Edit: It is now possible to install karaf on debian without building it yourself, the package installed is not the one described here, but the new and improved package built from source with native debian packaging tools, that can be found here  https://github.com/steinarb/karaf-debian

Apache karaf is an OSGi container/application server with some nice properties:

  1. It has an SSH server you can log into and a command line where you can inspect and configure the karaf instance
  2. It can be provisioned using apache maven, basically you can start with an empty karaf, ssh into the SSH server and pull in and start your application using “maven magic”
  3. It is much simpler to get an OSGi application up and running in apache karaf, than in any other system I have tried, since I first was to introduced to OSGi in 2006
  4. Karaf can also be used to run non-OSGi applications packaged as jar or war files
  5. In a development setting is very simple to deploy new versions of the code using maven and remote debug the deployed code frome eclipse or IntelliJ

Running karaf on a debian GNU/linux system is a little hampered by there not being a native .deb package. I have opened an RFP (Request For Packaging) bug for karaf in the debian bug tracker. When/if that issue is ever resolved as done, karaf will be easily availabel on debian and also on all of the distros that are based on debian (e.g. ubuntu and mint).

Until then do my own debian packaging. I forked the packaging I found at https://github.com/DemisR/karaf-deb-packaging and made some changes:

  1. Switched from oracle JDK 8, to openjdk 8
  2. Updated to karaf version 4.0.7 (the currently newest stable release at the time of forking), later upgraded to karaf 4.1.1 and again upgraded to karaf 4.1.2
  3. Use /var/lib/karaf/data instead of /usr/local/karaf/data
  4. Use package version “-1” instead of “-3”
  5. Switched from using the the service wrapper (karaf-wrapper) to plain systemd start using the scripts and config from bin/contrib in the karaf distribution
  6. Made the stop of running services more robust

The resulting .deb package will follow the usual service pattern of a debian service: the service will run with a user named after the service (i.e. user “karaf” which is the single member of group “karaf” and the owner of all files the service need to touch). The service will log to the regular debian syslog. The configuration will end up in /etc/karaf and all files not part of the installation will be left untouched on a .deb package uninstall and upgrade.

My fork of the packaging, lives at https://github.com/steinarb/karaf-deb-packaging

To create the package and install karaf, do the following steps:

  1. Log in as root on a debian system
  2. Install the prequisites for building the package, debian packages and ruby gem:
    apt-get update
    apt-get install git maven openjdk-8-jdk postgresql ruby ruby-dev build-essential
    gem install fpm
  3. Clone the packaging project and build the deb package:
    cd /tmp
    git clone https://github.com/steinarb/karaf-deb-packaging
    cd karaf-deb-packaging
    ./dist_karaf.sh
    mkdir -p /root/debs
    cp *.deb /root/debs
  4. Install the .deb package:
    dpkg --install karaf_4.1.4-1_all.deb

After karaf has been installed it is possible to log in as user “karaf”, with the following command

ssh -p 8101 karaf@localhost

The password is also “karaf” (without the quotes).

This opens the karaf console command line

        __ __                  ____
       / //_/____ __________ _/ __/
      / ,<  / __ `/ ___/ __ `/ /_
     / /| |/ /_/ / /  / /_/ / __/
    /_/ |_|\__,_/_/   \__,_/_/

  Apache Karaf (4.1.4)

Hit '<tab>' for a list of available commands
and '[cmd] --help' for help on a specific command.
Hit 'system:shutdown' to shutdown Karaf.
Hit '<ctrl-d>' or type 'logout' to disconnect shell from current session.

karaf@root()>

At this command line, you can eg.

  1. install an application
  2. start, stop and list running applications
  3. set the configuration used by the applications

But all of these except for the first, will be items for later posts.

Sign nginx website and dovecot imap server on debian with let’s encrypt

If you have a setup with a single server with multiple services (web, IMAP etc.), and one CNAME per service (www.somedomain.com, imap.somedomain.com), and you would like to get the services signed in a manner that doesn’t give warnings or errors in browsers  (especially browsers in phones and tablets with iOS and Android), then this article may be of interest.

Self-signed certificates is a nuisance and the cacert.org initiative has been losing support. Let’s encrypt offers the possibility of having free (as in both cost and feedom) SSL certificates that don’t give warnings in web browsers. The only problem the threshold of taking the time to figure out how to use it.

It turned out there wasn’t much figuring to: on a debian jessie GNU/linux system, the certbot program from eff.org takes care of everything, including keeping the certificates automatically updated (the .deb package for certbot sets up a cronjob that does the right thing).

The way certbot works is that it requires that each server you wish to sign must be accessible on http (port 80) and the local path “/.well-known/” on each server must be accessible and map to a file area that certbot can put files in.

The certbot program works by contacting let’s encrypt saying that it wants a certificate for a DNS name,  and let’s encrypt will then access the HTTP URL to verify that certbot is indeed running on a server that can be found using that DNS name.

This means that, for certbot to work:

  1. Even if your HTTP server responds only on HTTPS and/or requires authentication, you will need to make a plain HTTP connection available and have the local path “/.well-known/” map to a part of the file system, and be available without authentication
  2. Even if you’re making a certificate for a non-HTTP service (e.g. an IMAP server), you will need to make a plain http (port 80) server responding to that DNS CNAME that can serve the local parth “/.well-known/” from the local

This article explains how to set up free SSL certificates signed with let’s encrypt on an nginx web server and a dovecot IMAP server, on a debian jessie GNU/linux system.

The certbot instructions takes you part of the way, but it has some holes and not a lot of explanation, which is why I wrote this article.

The steps are:

  1. Add jessie-backports to APT (click the link and follow the instructions)
  2. Install certbot from jessie-backports:
    1. Open a command shell as root and give the following command:
      apt-get install certbot -t jessie-backports
      
  3. Disable the default nginx site
    1. Edit the /etc/nginx/sites-available/default file to have the following content:
      server {
              listen 80 default_server;
              listen [::]:80 default_server;
      
              root /var/www/html;
      
              server_name _;
      
              location / {
                      deny all;
              }
      }
      
    2. Run the following command in the command shell openes as root
      systemctl reload nginx
      
  4. Add DNS CNAME-records for the virtual hosts you are going to sign.
    In the examples used in this procedure, the host is hostname.somedomain.com and it has two CNAME aliases: http://www.somedomain.com and imap.somedomain.com.
  5. Add a http://www.mydomain.com nginx site
    1. Create a file /etc/nginx/available-sites/www with the following contents:
      server {
              listen 80;
              listen [::]:80;
      
              server_name www.mydomain.com;
      
              root /var/www/html;
      
              index index.html index.htm index.nginx-debian.html;
      
              location / {
                      allow all;
              }
      }
      
    2. Give the following commands in the command shell opened as root:
      cd /etc/nginx/enabled-sites/
      ln -s /etc/nginx/available-sites/www .
      systemctl reload nginx
      
  6. Add an imap.mydomain.com nginx site
    Note! This isn’t a real website but it is necessary to give HTTP access to a web server listening to this CNAME alias so that the certbot program can create and auto-update the certificate that dovecot uses.

    1. Create a file /etc/nginx/available-sites/imap with the following contents:
      # The port 80 listener only gives access to certbot
      server {
              listen 80;
              listen [::]:80;
      
              server_name imap.bang.priv.no;
      
              root /var/www-imap/;
      
              location /.well-known/ {
                      allow all;
              }
      
              location / {
                      deny all;
              }
      }
      
    2. Give the following commands in the command shell opened as root:
      cd /etc/nginx/enabled-sites/
      ln -s /etc/nginx/available-sites/imap .
      systemctl reload nginx
      
  7. Add a certificate for http://www.mydomain.com
    1. Give the following command in the command shell opened as root:
      certbot certonly --webroot -w /var/www/html -d www.mydomain.com
      
  8. Configure certificates for the http://www.mydomain.com nginx web site
    1. Change the /etc/nginx/available-sites/www file to the following:
      server {
              listen 80;
              listen [::]:80;
      
              server_name www.mydomain.com;
      
              # SSL configuration
              #
              listen 443 ssl default_server;
              listen [::]:443 ssl default_server;
              ssl_certificate     /etc/letsencrypt/live/www.mydomain.com/fullchain.pem;
              ssl_certificate_key /etc/letsencrypt/live/www.mydomain.com/privkey.pem;
      
              root /var/www/html;
      
              location / {
                      allow all;
              }
      }
      
    2. Give the following command in the command shell opened as root
      certbot certonly --webroot -w /var/www-imap -d imap.mydomain.com
      
    3. Open the https://www.somedomain.com server (replace with your actual URL) and observe that the browser reports it as secure with a valid certificate
  9. Add a certificate for imap.mydomain.com
    1. Give the following command in the command shell opened as root:
      certbot certonly --webroot -w /var/www-imap -d imap.mydomain.com
      
  10. Configure dovecot to use the imap.mydomain.com certificate
    1. Change/modify the following lines in the /etc/dovecot/conf.d/10-ssl.conf file:
      # SSL/TLS support: yes, no, required. <doc/wiki/SSL.txt>
      ssl = yes
      
      # PEM encoded X.509 SSL/TLS certificate and private key. They're opened before
      # dropping root privileges, so keep the key file unreadable by anyone but
      # root. Included doc/mkcert.sh can be used to easily generate self-signed
      # certificate, just make sure to update the domains in dovecot-openssl.cnf
      ssl_cert = </etc/letsencrypt/live/imap.bang.priv.no/fullchain.pem
      ssl_key = </etc/letsencrypt/live/imap.bang.priv.no/privkey.pem
      
    2. Give the following command in the command shell opened as root:
      /etc/init.d/dovecot reload
      

The certificates have a 90 day lifetime, but as mentioned earlier, the certificates will be automatically updated by certbot when they have 30 days valid time remaining. The certbot deb package installs a cronjob that runs twice every day at random second in the hour following 00:00 and 12:00 and checks if certificates needs to be updated and updates the ones that are ready for updating.