Frequently Asked Questions (FAQ)

Usage

What do I have to know about upgrading osm2pgsql to a newer version?

We are trying hard to make newer versions of osm2pgsql backwards compatible with older versions. Usually, you just have to upgrade the osm2pgsql executable and you are done. But there are cases where you have to wipe your database and start from scratch. See the release notes and the Upgrading appendix in the manual for details.

Can I have my tables in a specific database schema?

This is supported starting from version 1.4.0.

• For tables created by the flex output use the schema option on the table creation commands in the Lua config file.
• For tables created by the pgsql output use --output-pgsql-schema=SCHEMA.
• For middle tables use --middle-schema=SCHEMA.

Problems

Why is osm2pgsql so slow?

Osm2pgsql can take quite some time to import OSM data into a database, possibly many hours or even days. There are many reasons for this:

• There is a lot of data. A full planet will create a database with hundreds of GBytes, this just needs time.
• Your hardware may be underpowered. You should use fast SSDs and have plenty of RAM, for a planet import 64 GB RAM are pretty much the minimum, 128 GB are better.
• You have to tune your PostgreSQL config before using osm2pgsql. The default settings for PostgreSQL on most systems are totally wrong for a large database. Don’t forget to restart the database after tuning. See the manual for details.
• Compile osm2pgsql with the Lua JIT library for about 10 to 15 % speedup.
• If you don’t need your database to be updateable and you have enough RAM, use non-slim mode. Use a flat node file in slim mode unless you only import a small extract of the OSM data.
• If you have the choice, use input data in the PBF format, in some cases this can half the time needed for reading the input! Overall this can save up to 10% of the run-time of osm2pgsql.
• In some cases it makes sense to filter data before importing, see the manual.
• Make sure you are running the newest released version of osm2pgsql, there have been quite some performance improvements in the latest versions.

All that being said, on a reasonably modern machine with 128 GB RAM and SSDs you should be able to import a planet file in something like half a day. For some data about how long imports usually take, see Appendix E in the manual.

Why is an index not being built and there is no error message?

You are probably using a version of osm2pgsql before 1.3.0 which had a bug where errors happening while creating an index or certain other database operations were not reported but silently ignored. This is fixed in version 1.3.0.

You might be able to find information about the problem in the PostgreSQL log, most likely you ran out of disk space.

You should upgrade osm2pgsql to a current version.

Why is osm2pgsql crashing without reporting any useful error message?

This is most likely because you are running out of memory. Due to the way Linux systems “overcommit” memory, osm2pgsql can not detect that it is running out of memory, so it can’t tell you what’s going on.

Please read the Notes on Memory Usage in the manual to get some ideas how to handle this.

Why can’t I access a table column created by osm2pgsql?

Osm2pgsql usually creates table columns in your database that are named after the OSM tag used, for instance, the name tag might end up in a column called name. This can result in problems when the tag contains unusual characters, for instance the tag addr:city contains the colon character. Another problem are tags using a reserved name in the PostgreSQL database, for instance natural.

These names are allowed in PostgreSQL, but they need to be quoted with double quotes ("..."). Osm2pgsql does this quoting, so it doesn’t have any problem with these. But not all software does this.

You can define in the config file which columns you want, and, if you are using the flex output, decide on how exactly your columns should be named and used. To avoid this problem, you can, for instance, create a column named addr_city and fill it with the value of the addr:city tag.

Where is the coastline data?

In the pgsql output the natural=coastline tag is suppressed by default, even if you import the natural=* key. The main mapnik map renders coastlines from other sources so it does not need them. You can use the --keep-coastlines parameter to change this behavior if you want coastlines in your database.

See the Coastline Processing section in the manual

Why is this OSM object missing in my database?

There are many reasons why an OSM object might not end up in the database. First make sure it is actually in the input data. For ways and relations make sure the nodes or member objects they reference are in the input data. (You can use osmium check-refs for this.)

If the data is there it can still mean it is invalid in some way. Osm2pgsql has to build the geometry and if this fails, the object is silently ignored. This most often happens for multipolygon relations but can happen for other data, too.

There are several quality assurance tools out there that can help you with diagnosing issues like this. The OSM Inspector “Areas” view helps specifically with finding multipolygon problems.

Why do I get a duplicate ID error?

On current osm2pgsql the error looks something like this: ERROR: Input data is not ordered: node id 12815112 appears more than once. Older versions have different error messages.

This usually happens when reading an .osc.gz change file from planet.osm.org. You need to simplify the change file first. See the manual for more information.

Can osm2pgsql resume after a failed import?

Sometimes an osm2pgsql import fails, because of a bug in osm2pgsql or because you don’t have enough memory or because the computer crashes for unrelated reasons. Because the import can take many hours it would be great if osm2pgsql could resume the half-done import. Unfortunately this is currently not possible.

Error Messages

Here are some of the error messages osm2pgsql will produce and what you do if you get them.

bad_alloc or segmentation fault

Most likely this means that you ran out of memory. But it can also mean there is a bug in osm2pgsql. Check that you have enough memory, try with more. If the problem persists and you believe you have enough memory for what you are doing, report it.

Please read the Notes on Memory Usage in the manual to get some ideas how to handle this.

No tables defined in Lua config. Nothing to do!

You are using the flex output and the Lua config you have specified doesn’t define any output tables. You need to define at least one output table.

You’ll also get this error if you are using a pgsql Lua transform file with the flex output!

Features

Can we have support for other databases?

Osm2pgsql is tied closely to the PostgreSQL database and uses many special features of the PostgreSQL/PostGIS combination. Making sure osm2pgsql works with all PostgreSQL/PostGIS versions and has acceptable performance is already a big task. Adding support for other databases is not on our list of things to do.

We will consider pull requests with changes that make use of osm2pgsql with other databases easier (or possible) if the changes do not affect usability or performance for PostgreSQL users and if the changes can be cleanly integrated into the code. Because we don’t have the time to test new versions of osm2pgsql with other databases (or can’t even do it because we don’t have access to those databases), we can not guarantee that things will not break in the future.