Stackdump – an offline browser for StackExchange sites.

Stackdump was conceived for those who work in environments that do not have easy access to the StackExchange family of websites. It allows you to host a read-only instance of the StackExchange sites locally, accessible via a web browser.

Stackdump comprises of two components – the search indexer (Apache Solr) and the web application. It uses the StackExchange Data Dumps, published quarterly by StackExchange, as its source of data.


Stackdump home
Stackdump search results
Stackdump question view

System Requirements

Stackdump was written in Python and requires Python 2.5 or later (but not Python 3). There have been some reported issues with older versions of Python, so the ideal version to run is v2.7.6 (the latest 2.x version, as of writing). Stackdump also leverages Apache Solr, which requires the Java runtime (JRE), version 6 or later.

Besides that, there are no OS-dependent dependencies and should work on any platform that Python and Java run on (although it only comes bundled with Linux scripts at the moment). It was, however, developed and tested on CentOS 5 running Python 2.7 and JRE 6 update 27.

You will also need 7-zip to extract the data dump files, but Stackdump does not use it directly so you can perform the extraction on another machine first.

The amount of memory required for Stackdump depends on which dataset you want to import. For most datasets, at least 3GB of RAM is preferable. If you want to import StackOverflow, you must use a 64-bit operating system and a 64-bit version of Python, and also have at least 6GB of RAM available (or swap). If you do not have enough RAM available, the import process will likely fail with a MemoryError message at some point.

Make sure you have enough disk space too – having at least roughly the space used by the raw, extracted, data dump XML files available is a good rule of thumb (note that once imported, the raw data dump XML files are not needed by Stackdump any more).

Finally, Stackdump has been tested and works in the latest browsers (IE9, FF10+, Chrome, Safari). It degrades fairly gracefully in older browsers, although some will have rendering issues, e.g. IE8.

Changes and upgrading to v1.1

Version 1.1 fixes a few bugs, the major one being the inability to import the 2013 data dumps due to changes in the case of the filenames. It also adds a couple of minor features, including support for resolving and rewriting short question and answer permalinks.

Because changes have been made to the search schema and the search indexer has been upgraded (to Solr 4.5), all data will need to be re-indexed. Therefore there is no upgrade path; follow the instructions below to set up Stackdump again. It is recommended to install this new version in a new directory, instead of overwriting the existing one.

Changes and upgrading from v1.1 to v1.2

The major change in the v1.2 release are improvements to the speed of importing data. There are some other smaller changes, including new PowerShell scripts to start and manage Stackdump on Windows as well as a few bug fixes when running on Windows. The search indexing side of things has not changed, therefore data imported using v1.1 will continue to work in v1.2. Data from older versions however, needs to be re-indexed. See the above section on upgrading to v1.1 for more details.

Changes and upgrading from v1.2 to v1.3

v1.3 is primarily bugfix release, for a fairly serious bug. It turns out Stackdump has been subtly overwriting questions as more sites are imported because it assumed post IDs were unique across all sites, when they in fact were not. This meant as more sites were imported, the previous sites started to lose questions. The fix required a change to search index, therefore the data directory will need to be deleted and all data will need to be re-imported after installing this version. Thanks to yammesicka for reporting the issue.

Other changes include a new setting to allow disabling the link and image URL rewriting, and a change to the import_site command so it doesn’t bail immediately if there is a Solr connection issue – it will prompt and allow resumption after the connection issue has been resolved.

Importing the StackOverflow data dump, September 2013

The StackOverflow data dump has grown significantly since I started this project back in 2011. With the improvements in v1.2, on a VM with two cores and 4GB of RAM running CentOS 5.7 on a single, standard hard drive containing spinning pieces of metal,

  • it took 84719.565491 seconds to import it, or 23 hours, 31 minutes and 59.565491 seconds
  • once completed, it used up 20GB of disk space
  • during the import, roughly 30GB of disk space was needed
  • the import process used, at max, around 2GB of RAM.

In total, the StackOverflow data dump has 15,933,529 posts (questions and answers), 2,332,403 users and a very large number of comments.

I attempted this on a similarly spec’ed Windows 7 64-bit VM as well – 23 hours later and it is still trying to process the comments. The SQLite, Python or just disk performance is very poor for some reason. Therefore, if you intend on importing StackOverflow, I would advise you to run Stackdump on Linux instead. The smaller sites all complete without a reasonable time though, and there are no perceptible issues with performance as far as I’m aware on Windows.

Reports on importing the StackOverflow data dump, September 2014

Due to the growth of the dataset, the import process now requires at least 6GB of RAM. This also means you must use a 64-bit operating system and a 64-bit version of Python.

Setting up

Stackdump was designed for offline environments or environments with poor internet access, therefore it is bundled with all the dependencies it requires (with the exception of Python, Java and 7-zip).

As long as you have:

  • Python, version 2.5 or later but not version 3 (tested with v2.7.6),
  • Java, version 6 (1.6) or later,
  • Stackdump,
  • the StackExchange Data Dump (download the sites you wish to import – note that StackOverflow is split into 7 archive files; only Comments, Posts and Users are required but after extraction the files need to be renamed to Comments.xml, Posts.xml and Users.xml respectively), and
  • 7-zip (needed to extract the data dump files)

...you should be able to get an instance up and running.

If you are using a 64-bit operating system, get the 64-bit version of Python.

To provide a better experience, Stackdump can use the RSS feed content to pre-fill some of the required details during the import process, as well as to display the site logos in the app. Stackdump comes bundled with a script that downloads and places these bits in the right places. If you’re in a completely offline environment however, it may be worth running this script on a connected box first.

Windows users

If you’re using Windows, you will need to substitute the appropriate PowerShell equivalent command for the Stackdump scripts used below. These equivalent PowerShell scripts are in the Stackdump root directory, alongside their Unix counterparts. The names are roughly the same, with the exception of manage.sh, which in PowerShell has been broken up into two scripts, List-StackdumpCommands.ps1 and Run-StackdumpCommand.ps1.

Remember to set your PowerShell execution policy to at least RemoteSigned first as these scripts are not signed. Use the Get-ExecutionPolicy cmdlet to see the current policy, and Set-ExecutionPolicy to set it. You will need to have administrative privileges to set it.

Extract Stackdump

Stackdump was designed to be self-contained, so to get it up and running, simply extract the Stackdump download archive to an appropriate location.

Verify dependencies

Next, you should verify that the required Java and Python versions are accessible in the PATH. (If you haven’t installed them yet, now is a good time to do so.)

Type java -version and check that it is at least version 1.6.

If you’re using Java 7 on Linux and you see an error similar to the following – Error: failed /opt/jre1.7.0_40/lib/i386/server/libjvm.so, because /opt/jre1.7.0_40/lib/i386/server/libjvm.so: cannot restore segment prot after reloc: Permission denied
this is because you have SELinux enabled. You will need to tell SELinux to allow Java to run by using the following command as root (amending the path as necessary) – chcon -t textrel_shlib_t /opt/jre1.7.0_40/lib/i386/server/libjvm.so

Then type python -V and check that it is version 2.5 or later (and not Python 3). Ideally this should be v2.7.6 or later as there have been some reported issues with earlier versions.

If you would rather not put these versions in the PATH (e.g. you don’t want to override the default version of Python in your Linux distribution), you can tell Stackdump which Java and/or Python to use explicitly by creating a file named JAVA_CMD or PYTHON_CMD respectively in the Stackdump root directory, and placing the path to the executable in there.

Download additional site information

As mentioned earlier, Stackdump can use additional information available in the StackExchange RSS feed to pre-fill required details during the site import process and to show the logos for each site.

To start the download, execute the following command in the Stackdump root directory –

./manage.sh download_site_info

If Stackdump will be running in a completely offline environment, it is recommended that you extract and run this command in a connected environment first. If that is not possible, you can manually download the required pieces –

  • download the RSS feed to a file
  • for each site you will be importing, work out the site key and download the logo by substituting the site key into this URL: http://sstatic.net/site_key/img/icon-48.png where site_key is the site key. The site key is generally the bit in the URL before .stackexchange.com, or just the domain without the TLD, e.g. for the Salesforce StackExchange at http://salesforce.stackexchange.com, it is just salesforce, while for Server Fault at http://serverfault.com, it is serverfault.

The RSS feed file should be copied to the file stackdump_dir/data/sites (create the data directory if it doesn’t exist), and the logos should be copied to the stackdump_dir/python/media/images/logos directory and named with the site key and file type extension, e.g. serverfault.png.

Import sites

Each data dump for a StackExchange site is a 7-zip file. Extract the file corresponding to the site you wish to import into a temporary directory. It should have a bunch of XML files in it when complete.

Now make sure you have the search indexer up and running. This can be done by simply executing the stackdump_dir/start_solr.sh command.

To start the import process, execute the following command –

stackdump_dir/manage.sh import_site --base-url site_url --dump-date dump_date path_to_xml_files

... where site_url is the URL of the site you’re importing, e.g. android.stackexchange.com; dump_date is the date of the data dump you’re importing, e.g. August 2012, and finally path_to_xml_files is the path to the directory containing the XML files that were just extracted. The dump_date is a text string that is shown in the app only, so it can be in any format you want.

For example, to import the August 2012 data dump of the Android StackExchange site, with the files extracted into /tmp/android, you would execute –

stackdump_dir/manage.sh import_site --base-url android.stackexchange.com --dump-date "August 2012" /tmp/android

It is normal to get messages about unknown PostTypeIds and missing comments and answers. These errors are likely due to those posts being hidden via moderation.

This can take anywhere between a minute to 20 hours or more depending on the site you’re importing. As a rough guide, android.stackexchange.com took a minute on my VM, while stackoverflow.com took just under 24 hours.

Repeat these steps for each site you wish to import. Do not attempt to import multiple sites at the same time; it will not work and you may end up with half-imported sites.

The import process can be cancelled at any time without any adverse effect, however on the next run it will have to start from scratch again.

Start the app

To start Stackdump, execute the following command –


... and visit port 8080 on that machine. That’s it – your own offline, read-only instance of StackExchange.

If you need to change the port that it runs on, or modify other settings that control how Stackdump works; see the 'Optional configuration’ section below for more details.

Both the search indexer and the app need to be running for Stackdump to work.

Optional configuration

There are a few settings for those who like to tweak. There’s no need to adjust them normally though; the default settings should be fine.

The settings file is located in stackdump_dir/python/src/stackdump/settings.py. The web component will need to be restarted after changes have been made for them to take effect.

  • SERVER_HOST – the network interface to run the Stackdump web app on. Use '’ for all interfaces, or '’ for localhost only. By default, it runs on all interfaces.
  • SERVER_PORT – the port to run the Stackdump web app on. The default port is 8080.
  • SOLR_URL – the URL to the Solr instance. The default assumes Solr is running on the same system. Change this if Solr is running on a different system.
  • NUM_OF_DEFAULT_COMMENTS – the number of comments shown by default for questions and answers before the remaining comments are hidden (and shown when clicked). The default is 3 comments.
  • NUM_OF_RANDOM_QUESTIONS – the number of random questions shown on the home page of Stackdump and the site pages. The default is 3 questions.
  • REWRITE_LINKS_AND_IMAGES – by default, all links are rewritten to either point internally or be marked as an external link, and image URLs are rewritten to point to a placeholder image. Set this setting to False to disable this behaviour.

Running Stackdump as a service

Stackdump comes bundled with some init.d scripts as well which were tested on CentOS 5. These are located in the init.d directory. To use these, you will need to modify them to specify the path to the Stackdump root directory and the user to run under.

Another option is to use Supervisor with a simple configuration file, e.g.,



Supervisor v3.0b1 or later is required, due to the stopasgroup parameter. Without this parameter, Supervisor will not be able to stop the Stackdump components properly as they’re being executed from a script.

Yet another option for those using newer Linux distributions is to create native systemd service definitions of type simple for each of the components.


Stackdump stores all its data in the data directory under its root directory. If you want to start fresh, just stop the app and the search indexer, delete that directory and restart the app and search indexer.

To delete certain sites from Stackdump, use the manage_sites management command –

stackdump_dir/manage.sh manage_sites -l to list the sites (and their site keys) currently in the system;
stackdump_dir/manage.sh manage_sites -d site_key to delete a particular site.

It is not necessary to delete a site before importing a new data dump of it though; the import process will automatically purge the old copy during the import process.


Stackdump leverages several open-source projects to do various things, including –

Things not supported… yet

  • searching or browsing by tags
  • tag wiki pages
  • badges
  • post history, e.g. reasons why are a post was closed are not listed


Stackdump is licensed under the MIT License.