Elasticsearch integration

Introduced in GitLab Starter 8.4. Support for Amazon Elasticsearch was introduced in GitLab Starter 9.0.

This document describes how to set up Elasticsearch with GitLab. Once enabled, you’ll have the benefit of fast search response times and the advantage of two special searches:

Version Requirements

GitLab version Elasticsearch version
GitLab Enterprise Edition 8.4 - 8.17 Elasticsearch 2.4 with Delete By Query Plugin installed
GitLab Enterprise Edition 9.0 - 11.4 Elasticsearch 5.1 - 5.5
GitLab Enterprise Edition 11.5+ Elasticsearch 5.6 - 6.x

Installing Elasticsearch

Elasticsearch is not included in the Omnibus packages. You will have to install it yourself whether you are using the Omnibus package or installed GitLab from source. Providing detailed information on installing Elasticsearch is out of the scope of this document.

Once the data is added to the database or repository and Elasticsearch is enabled in the admin area the search index will be updated automatically. Elasticsearch can be installed on the same machine as GitLab, or on a separate server, or you can use the Amazon Elasticsearch service.

You can follow the steps as described in the official web site or use the packages that are available for your OS.

Elasticsearch repository indexer (beta)

In order to improve elasticsearch indexing performance GitLab has made available a new indexer written in Go. This will replace the included Ruby indexer in the future but should be considered beta software for now, so there may be some bugs.

If you would like to use it, please follow the instructions below.

Installation

First, we need to install some dependencies, then we’ll build and install the indexer itself.

Dependencies

This project relies on ICU for text encoding, therefore we need to ensure the development packages for your platform are installed before running make.

Debian / Ubuntu

To install on Debian or Ubutu, run:

sudo apt install libicu-dev
Mac OSX

To install on macOS, run:

brew install icu4c
export PKG_CONFIG_PATH="/usr/local/opt/icu4c/lib/pkgconfig:$PKG_CONFIG_PATH"

Building and installing

To build and install the indexer, run:

make
sudo make install

The gitlab-elasticsearch-indexer will be installed to /usr/local/bin.

You can change the installation path with the PREFIX env variable. Please remember to pass the -E flag to sudo if you do so.

Example:

PREFIX=/usr sudo -E make install

Once installed, enable it under your instance’s elasticsearch settings explained below.

System Requirements

Elasticsearch requires additional resources in excess of those documented in the GitLab system requirements. These will vary by installation size, but you should ensure at least an additional 8 GiB of RAM for each Elasticsearch node, per the official guidelines.

Storage requirements also vary based on installation side, but as a rule of thumb, you should allocate the total size of your production database, plus two-thirds of the total size of your git repositories. Efforts to reduce this total are being tracked in this epic: gitlab-org&153.

Enabling Elasticsearch

In order to enable Elasticsearch, you need to have admin access. Go to Admin > Settings > Integrations and find the “Elasticsearch” section.

The following Elasticsearch settings are available:

Parameter Description
Elasticsearch indexing Enables/disables Elasticsearch indexing. You may want to enable indexing but disable search in order to give the index time to be fully completed, for example. Also keep in mind that this option doesn’t have any impact on existing data, this only enables/disables background indexer which tracks data changes. So by enabling this you will not get your existing data indexed, use special rake task for that as explained in Adding GitLab’s data to the Elasticsearch index.
Use the new repository indexer (beta) Perform repository indexing using GitLab Elasticsearch Indexer.
Search with Elasticsearch enabled Enables/disables using Elasticsearch in search.
URL The URL to use for connecting to Elasticsearch. Use a comma-separated list to support clustering (e.g., “http://host1, https://host2:9200"). If your Elasticsearch instance is password protected, pass the username:password in the URL (e.g., http://<username>:<password>@<elastic_host>:9200/).
Limit namespaces and projects that can be indexed Enabling this will allow you to select namespaces and projects to index. All other namespaces and projects will use database search instead. Please note that if you enable this option but do not select any namespaces or projects, none will be indexed. Read more below.
Using AWS hosted Elasticsearch with IAM credentials Sign your Elasticsearch requests using AWS IAM authorization or AWS EC2 Instance Profile Credentials. The policies must be configured to allow es:* actions.
AWS Region The AWS region your Elasticsearch service is located in.
AWS Access Key The AWS access key.
AWS Secret Access Key The AWS secret access key.

Limiting namespaces and projects

If you select Limit namespaces and projects that can be indexed, more options will become available limit namespaces and projects options

You can select namespaces and projects to index exclusively. Please note that if the namespace is a group it will include any sub-groups and projects belonging to those sub-groups to be indexed as well.

You can filter the selection dropdown by writing part of the namespace or project name you’re interested in. limit namespace filter

Note: If no namespaces or projects are selected, no Elasticsearch indexing will take place.
Warning: If you have already indexed your instance, you will have to regenerate the index in order to delete all existing data for filtering to work correctly. To do this run the rake tasks gitlab:elastic:create_empty_index and gitlab:elastic:clear_index_status Afterwards, removing a namespace or a projeect from the list will delete the data from the Elasticsearch index as expected.

Disabling Elasticsearch

To disable the Elasticsearch integration:

  1. Navigate to the Admin > Settings > Integrations
  2. Find the ‘Elasticsearch’ section and uncheck ‘Search with Elasticsearch enabled’ and ‘Elasticsearch indexing’
  3. Click Save for the changes to take effect

Adding GitLab’s data to the Elasticsearch index

Indexing small instances (database size less than 500 MiB, size of repos less than 5 GiB)

Configure Elasticsearch’s host and port in Admin > Settings > Integrations. Then create empty indexes using one of the following commands:

# Omnibus installations
sudo gitlab-rake gitlab:elastic:create_empty_index

# Installations from source
bundle exec rake gitlab:elastic:create_empty_index RAILS_ENV=production

Then enable Elasticsearch indexing and run repository indexing tasks:

# Omnibus installations
sudo gitlab-rake gitlab:elastic:index

# Installations from source
bundle exec rake gitlab:elastic:index RAILS_ENV=production

Enable Elasticsearch search.

Indexing large instances

Note: After indexing the repositories asynchronously, you MUST index the database to be able to search.

Configure Elasticsearch’s host and port in Admin > Settings > Integrations. Then create empty indexes using one of the following commands:

# Omnibus installations
sudo gitlab-rake gitlab:elastic:create_empty_index

# Installations from source
bundle exec rake gitlab:elastic:create_empty_index RAILS_ENV=production

Indexing large Git repositories can take a while. To speed up the process, you can temporarily disable auto-refreshing and replicating. In our experience you can expect a 20% time drop. We’ll enable them when indexing is done. This step is optional!

curl --request PUT localhost:9200/gitlab-production/_settings --data '{
    "index" : {
        "refresh_interval" : "-1",
        "number_of_replicas" : 0
    } }'

Then enable Elasticsearch indexing and run repository indexing tasks:

# Omnibus installations
sudo gitlab-rake gitlab:elastic:index_repositories_async

# Installations from source
bundle exec rake gitlab:elastic:index_repositories_async RAILS_ENV=production

This enqueues a number of Sidekiq jobs to index your existing repositories. You can view the jobs in the admin panel (they are placed in the elastic_batch_project_indexer) queue), or you can query indexing status using a rake task:

# Omnibus installations
sudo gitlab-rake gitlab:elastic:index_repositories_status

# Installations from source
bundle exec rake gitlab:elastic:index_repositories_status RAILS_ENV=production

Indexing is 65.55% complete (6555/10000 projects)

By default, one job is created for every 300 projects. For large numbers of projects, you may wish to increase the batch size, by setting the BATCH environment variable.

You can also run the initial indexing synchronously - this is most useful if you have a small number of projects, or need finer-grained control over indexing than Sidekiq permits:

# Omnibus installations
sudo gitlab-rake gitlab:elastic:index_repositories

# Installations from source
bundle exec rake gitlab:elastic:index_repositories RAILS_ENV=production

It might take a while depending on how big your Git repositories are.

If you want to run several tasks in parallel (probably in separate terminal windows) you can provide the ID_FROM and ID_TO parameters:

# Omnibus installations
sudo gitlab-rake gitlab:elastic:index_repositories ID_FROM=1001 ID_TO=2000

# Installations from source
bundle exec rake gitlab:elastic:index_repositories ID_FROM=1001 ID_TO=2000 RAILS_ENV=production

Where ID_FROM and ID_TO are project IDs. Both parameters are optional. As an example, if you have 3,000 repositories and you want to run three separate indexing tasks, you might run:

# Omnibus installations
sudo gitlab-rake gitlab:elastic:index_repositories ID_TO=1000
sudo gitlab-rake gitlab:elastic:index_repositories ID_FROM=1001 ID_TO=2000
sudo gitlab-rake gitlab:elastic:index_repositories ID_FROM=2001

# Installations from source
bundle exec rake gitlab:elastic:index_repositories RAILS_ENV=production ID_TO=1000
bundle exec rake gitlab:elastic:index_repositories RAILS_ENV=production ID_FROM=1001 ID_TO=2000
bundle exec rake gitlab:elastic:index_repositories RAILS_ENV=production ID_FROM=2001

Sometimes your repository index process gitlab:elastic:index_repositories or gitlab:elastic:index_repositories_async can get interrupted. This may happen for many reasons, but it’s always safe to run the indexing job again - it will skip those repositories that have already been indexed.

As the indexer stores the last commit SHA of every indexed repository in the database, you can run the indexer with the special parameter UPDATE_INDEX and it will check every project repository again to make sure that every commit in that repository is indexed, it can be useful in case if your index is outdated:

# Omnibus installations
sudo gitlab-rake gitlab:elastic:index_repositories UPDATE_INDEX=true ID_TO=1000

# Installations from source
bundle exec rake gitlab:elastic:index_repositories UPDATE_INDEX=true ID_TO=1000 RAILS_ENV=production

You can also use the gitlab:elastic:clear_index_status Rake task to force the indexer to “forget” all progresss, so retrying the indexing process from the start.

To index all wikis:

# Omnibus installations
sudo gitlab-rake gitlab:elastic:index_wikis

# Installations from source
bundle exec rake gitlab:elastic:index_wikis RAILS_ENV=production

The wiki indexer also supports the ID_FROM and ID_TO parameters if you want to limit a project set.

Index all database entities (Keep in mind it can take a while so consider using screen or tmux):

# Omnibus installations
sudo gitlab-rake gitlab:elastic:index_database

# Installations from source
bundle exec rake gitlab:elastic:index_database RAILS_ENV=production

Enable replication and refreshing again after indexing (only if you previously disabled it):

curl --request PUT localhost:9200/gitlab-production/_settings --data '{
    "index" : {
        "number_of_replicas" : 1,
        "refresh_interval" : "1s"
    } }'

A force merge should be called after enabling the refreshing above:

curl --request POST 'http://localhost:9200/_forcemerge?max_num_segments=5'

Enable Elasticsearch search in Admin > Settings > Integrations. That’s it. Enjoy it!

Tuning

Deleted documents

Whenever a change or deletion is made to an indexed GitLab object (a merge request description is changed, a file is deleted from the master branch in a repository, a project is deleted, etc), a document in the index is deleted. However, since these are “soft” deletes, the overall number of “deleted documents”, and therefore wasted space, increases. Elasticsearch does intelligent merging of segments in order to remove these deleted documents. However, depending on the amount and type of activity in your GitLab installation, it’s possible to see as much as 50% wasted space in the index.

In general, we recommend simply letting Elasticseach merge and reclaim space automatically, with the default settings. From Lucene’s Handling of Deleted Documents, “Overall, besides perhaps decreasing the maximum segment size, it is best to leave Lucene’s defaults as-is and not fret too much about when deletes are reclaimed.”

However, some larger installations may wish to tune the merge policy settings:

  • Consider reducing the index.merge.policy.max_merged_segment size from the default 5 GB to maybe 2 GB or 3 GB. Merging only happens when a segment has at least 50% deletions. Smaller segment sizes will allow merging to happen more frequently.

    curl --request PUT http://localhost:9200/gitlab-production/_settings --data '{
      "index" : {
        "merge.policy.max_merged_segment": "2gb"
      }
    }'
    
  • You can also adjust index.merge.policy.reclaim_deletes_weight, which controls how aggressively deletions are targetd. But this can lead to costly merge decisions, so we recommend not changing this unless you understand the tradeoffs.

    curl --request PUT http://localhost:9200/gitlab-production/_settings --data '{
      "index" : {
        "merge.policy.reclaim_deletes_weight": "3.0"
      }
    }'
    
  • Do not do a force merge to remove deleted documents. A warning in the documentation states that this can lead to very large segments that may never get reclaimed, and can also cause significant performance or availability issues.

Troubleshooting

Here are some common pitfalls and how to overcome them:

  • I updated GitLab and now I can’t find anything

    We continuously make updates to our indexing strategies and aim to support newer versions of Elasticsearch. When indexing changes are made, it may be necessary for you to reindex after updating GitLab.

  • I indexed all the repositories but I can’t find anything

    Make sure you indexed all the database data as stated above.

  • I indexed all the repositories but then switched elastic search servers and now I can’t find anything

    You will need to re-run all the rake tasks to re-index the database, repositories, and wikis.

  • No new data is added to the Elasticsearch index when I push code

    When performing the initial indexing of blobs, we lock all projects until the project finishes indexing. It could happen that an error during the process causes one or multiple projects to remain locked. In order to unlock them, run the gitlab:elastic:clear_locked_projects rake task.

  • “Can’t specify parent if no parent field has been configured”

    If you enabled Elasticsearch before GitLab 8.12 and have not rebuilt indexes you will get exception in lots of different cases:

      Elasticsearch::Transport::Transport::Errors::BadRequest([400] {
          "error": {
              "root_cause": [{
                  "type": "illegal_argument_exception",
                  "reason": "Can't specify parent if no parent field has been configured"
              }],
              "type": "illegal_argument_exception",
              "reason": "Can't specify parent if no parent field has been configured"
          },
          "status": 400
      }):
    

    This is because we changed the index mapping in GitLab 8.12 and the old indexes should be removed and built from scratch again, see details in the 8-11-to-8-12 update guide.

  • Exception Elasticsearch::Transport::Transport::Errors::BadRequest

    If you have this exception (just like in the case above but the actual message is different) please check if you have the correct Elasticsearch version and you met the other requirements. There is also an easy way to check it automatically with sudo gitlab-rake gitlab:check command.

  • Exception Elasticsearch::Transport::Transport::Errors::RequestEntityTooLarge

      [413] {"Message":"Request size exceeded 10485760 bytes"}
    

    This exception is seen when your Elasticsearch cluster is configured to reject requests above a certain size (10MiB in this case). This corresponds to the http.max_content_length setting in elasticsearch.yml. Increase it to a larger size and restart your Elasticsearch cluster.

    AWS has fixed limits for this setting (“Maximum Size of HTTP Request Payloads”), based on the size of the underlying instance.