osmdbt icon indicating copy to clipboard operation
osmdbt copied to clipboard

OSM Database Replication Tools

OSM Database Tools

Tools for creating replication feeds from the main OSM database.

These tools are only useful if you run an OSM database like the main central OSM database!

Prerequisites

You need a C++17 compliant compiler. GCC 8 and later as well as clang 7 and later are known to work.

You also need the following libraries:

Libosmium (>= 2.15.0)
    https://osmcode.org/libosmium
    Debian/Ubuntu: libosmium2-dev
    Fedora/CentOS: libosmium-devel

Protozero (>= 1.6.3)
    https://github.com/mapbox/protozero
    Debian/Ubuntu: libprotozero-dev
    Fedora/CentOS: protozero-devel

boost-filesystem (>= 1.55)
    https://www.boost.org/doc/libs/1_55_0/libs/filesystem/doc/index.htm
    Debian/Ubuntu: libboost-filesystem-dev
    Fedora/CentOS: boost-devel

boost-program-options (>= 1.55)
    https://www.boost.org/doc/libs/1_55_0/doc/html/program_options.html
    Debian/Ubuntu: libboost-program-options-dev
    Fedora/CentOS: boost-devel

bz2lib
    http://www.bzip.org/
    Debian/Ubuntu: libbz2-dev
    Fedora/CentOS: bzip2-devel

zlib
    https://www.zlib.net/
    Debian/Ubuntu: zlib1g-dev
    Fedora/CentOS: zlib-devel

Expat
    https://libexpat.github.io/
    Debian/Ubuntu: libexpat1-dev
    Fedora/CentOS: expat-devel

cmake
    https://cmake.org/
    Debian/Ubuntu: cmake
    Fedora/CentOS: cmake

yaml-cpp
    https://github.com/jbeder/yaml-cpp
    Debian/Ubuntu: libyaml-cpp-dev
    Fedora/CentOS: yaml-cpp-devel

libpqxx (version 6)
    https://github.com/jtv/libpqxx/
    Debian/Ubuntu: libpqxx-dev
    Fedora/CentOS: libpqxx-devel

Pandoc
    (Needed to build documentation, optional)
    https://pandoc.org/
    Debian/Ubuntu: pandoc
    Fedora/CentOS: pandoc

gettext
    (envsubst command for tests)
    Debian/Ubuntu: gettext-base
    Fedora/CentOS: gettext

PostgreSQL database and pg_virtualenv
    Debian/Ubuntu: postgresql-common, postgresql-server-dev-all
    Fedora/CentOS: postgresql-server, postgresql-server-devel

On Linux systems most of these libraries are available through your package manager, see the list above for the names of the packages. But make sure to check the versions. If the packaged version available is not new enough, you'll have to install from source. Most likely this is the case for Libosmium.

Building

Use CMake to build in the usual way, for instance:

mkdir build
cd build
cmake ..
cmake --build .

If there are several versions of PostgreSQL installed on your system, you might have to set the PG_CONFIG variable to the full path like so:

cmake -DPG_CONFIG=/usr/lib/postgresql/14/bin/pg_config ..

If you don't want to build the PostgreSQL plugin set BUILD_PLUGIN to OFF.

cmake -DBUILD_PLUGIN=OFF ..

If you only want to build the PostgreSQL plugin:

cd postgresql-plugin
mkdir build
cd build
cmake ..
cmake --build .

Database Setup

You need a PostgreSQL database with

  • config option wal_level=logical,
  • config option max_replication_slots set to at least 1,
  • a user with REPLICATION attribute, and
  • a database containing an OSM database where this user has access.

There is an unofficial test/structure.sql provided in this repository to set up an OSM database for testing. Do not use it for production, use the official way of installing an OSM database instead.

(The original for this file is in the openstreetmap-website repository at https://github.com/openstreetmap/openstreetmap-website/raw/master/db/structure.sql)

Running

There are several commands in the build/src directory. They all can be called with -h or --help to see how they are run. They all need a common config file, a template is in osmdbt-config.yaml. This will be found automatically if it is in the current directory, use -c or --config to set a different path.

Documentation

There are man pages for all commands in man and an overview page in man/osmdbt.md.

If you have pandoc installed they will be built when running make.

Tests

To run the tests after build call ctest.

Debian Package

To create a Debian/Ubuntu package, call debuild -I.

The Debian package will contain the executables and the man pages. It will not contain the PostgreSQL plugin, because that needs to be built for the exact PostgreSQL version you have.

Usage

First set up the configuration file and make sure you can access the database by running:

osmdbt-testdb

Then enable replication:

osmdbt-enable-replication

After that get current log files once every minute (or whatever you want the update interval to be). Use cron or something like it to handle this:

osmdbt-get-log --catchup

To create an OSM change file from the log, call

osmdbt-create-diff

To disable replication, use:

osmdbt-disable-replication

For more details see the individual man pages.

How it works in detail

This section describes how everything is supposed to work in detail. The whole process is controlled from one shell script run once per minute. It looks something like this:

#!/bin/sh

set -e

osmdbt-catchup
osmdbt-get-log
# optionally copy log file(s) to other hosts
osmdbt-catchup
osmdbt-create-diff

1. Catch up old log files

If there are complete log files left over from a crash, they will be in the log_dir directory and named *.log.

osmdbt-catchup is called without command line arguments. It finds those left-over log files and tells the PostgreSQL database the largest of the LSNs so that the database can "forget" all changes before that.

If there was no crash, no such log files are found and osmdbt-catchup does nothing.

2. Create log file

Now osmdbt-get-log is called which creates a log file in the log_dir named something like osm-repl-2020-03-18T14:18:49Z-lsn-0-1924DE0.log. The file is first created with the suffix .new, synced to disk, then renamed and the directory is synced.

If any of these steps fail or if the host crashes, a .new file might be left around, which should be flagged for the sysadmin to take care of. The file can be removed without loosing data, but the circumstances should be reviewed in case there is some systematic problem.

3. Copy log file to separate host (optional)

All files named *.log in the log_dir can now be copied (using scp or rsync or so) to a separate host for safekeeping. These will only be used if the local host crashes and log files on its disk are lost. In this case manual intervention is necessary.

4. Catch up database to new log file

Now osmdbt-catchup is called to catch up the database to the log file just created in step 2.

If the system crashes in step 2, 3, or 4 a log file might be left around without the database being updated. In this case step 1 of the next cycle will pick this up and do the database update.

5. Creating diff file

Now osmdbt-create-diff is called which reads any log files in the log_dir and creates replication diff files. Files are first created in the tmp_dir directory and then moved into place in the changes_dir and its subdirectories. osmdbt-create-diff will also read the state.txt in the changes_dir file and create a new one. See the manual page for osmdbt-create-diff for the details on how this is done exactly.

Log files and lock files

  • The programs osmdbt-get-log, osmdbt-fake-log, and osmdbt-catchup use the same PID/lock file run_dir/osmdbt-log making sure that only one of them is running.
  • The program osmdbt-create-diff uses a different PID/lock file, it can run in parallel to the other programs, but only one copy of it will run.
  • osmdbt-create-diff can handle any number of log files, so if it is not run for a while it will recover by reading all log files it finds and creating one replication diff file with the (sorted) data from all of them.
  • All programs will write files under different names or in separate directories, sync the files and only then move them into place atomically. After that directories are synced.
  • After diff and state files are created in the tmp_dir, before they are moved to the changes_dir, a file osmdbt-create-diff.lock is created in tmp_dir. This is removed after osmdbt-create-diff finished moving all state and diff files to their final place and appending the .done suffix to the log file(s). If this is left lying around, osmdbt-create-diff will not run any more and the sysadmin needs to intervene.

External processing needed

When run in production you should regularly

  • remove old log files marked as done (files in log_dir named *.log.done)
  • remove log file copies you might have made on a separate host

To make sure everything runs smoothly, the age of the PID files can be checked (should never be more than a few seconds) and the existence of older (more than a minute or so) log files named *.log.new.

License

Copyright (C) 2021-2022 Jochen Topf ([email protected])

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see https://www.gnu.org/licenses/.

Authors

This program was written and is maintained by Jochen Topf ([email protected]).