info: To determine the technical writer assigned to the Stage/Group associated with this page, see https://about.gitlab.com/handbook/product/ux/technical-writing/#assignments
When we change Elasticsearch supported versions in GitLab, we announce them in [deprecation notes](https://about.gitlab.com/handbook/marketing/blog/release-posts/#deprecations) in monthly release posts
If you are using a compatible version and after connecting to OpenSearch, you get the message `Elasticsearch version not compatible`, [unpause indexing](#unpause-indexing).
## System requirements
Elasticsearch requires additional resources to those documented in the
[GitLab system requirements](../../install/requirements.md).
Memory, CPU, and storage resource amounts vary depending on the amount of data you index into the Elasticsearch cluster. Heavily used Elasticsearch clusters may require more resources. The [`estimate_cluster_size`](#gitlab-advanced-search-rake-tasks) Rake task ([introduced](https://gitlab.com/gitlab-org/gitlab/-/issues/221177) in GitLab 13.10) uses the total repository size to estimate the Advanced Search storage requirements.
source. You must [install it separately](https://www.elastic.co/guide/en/elasticsearch/reference/7.16/install-elasticsearch.html "Elasticsearch 7.x installation documentation") and ensure you select your version. Detailed information on how to install Elasticsearch is out of the scope of this page.
You can install Elasticsearch yourself, or use a cloud hosted offering such as [Elasticsearch Service](https://www.elastic.co/elasticsearch/service) (available on AWS, GCP, or Azure) or the [Amazon OpenSearch](https://docs.aws.amazon.com/opensearch-service/latest/developerguide/gsg.html)
service.
You should install Elasticsearch on a separate server. Running Elasticsearch on the same server as GitLab is not recommended and can cause a degradation in GitLab instance performance.
For a single node Elasticsearch cluster, the functional cluster health status is always yellow due to the allocation of the primary shard. Elasticsearch cannot assign replica shards to the same node as primary shards.
The search index updates after you:
- Add data to the database or repository.
- [Enable Elasticsearch](#enable-advanced-search) in the Admin Area.
## Upgrade to a new Elasticsearch major version
> - Elasticsearch 6.8 support is removed with GitLab 15.0.
> - Upgrading from GitLab 14.10 to 15.0 requires that you are using any version of Elasticsearch 7.x.
You are not required to change the GitLab configuration when you upgrade Elasticsearch.
## Elasticsearch repository indexer
To index Git repository data, GitLab uses an [indexer written in Go](https://gitlab.com/gitlab-org/gitlab-elasticsearch-indexer).
Depending on your GitLab version, there are different installation procedures for the Go indexer:
- For Omnibus GitLab 11.8 or greater, see [Omnibus GitLab](#omnibus-gitlab).
- For installations from source or older versions of Omnibus GitLab,
[install the indexer from source](#from-source).
- If you are using GitLab Development Kit, see [GDK Elasticsearch how-to](https://gitlab.com/gitlab-org/gitlab-development-kit/-/blob/main/doc/howto/elasticsearch.md).
### Omnibus GitLab
Starting with GitLab 11.8, the Go indexer is included in Omnibus GitLab.
The former Ruby-based indexer was removed in [GitLab 12.3](https://gitlab.com/gitlab-org/gitlab/-/issues/6481).
### From source
First, we need to install some dependencies, then we build and install
the [`elasticsearch.log`](../../administration/logs/index.md#elasticsearchlog) file and the [`sidekiq.log`](../../administration/logs/index.md#sidekiqlog) file with a `json.exception.class` of `Gitlab::Elastic::Indexer::Error`.
For GitLab instances with more than 50 GB repository data you can follow the instructions for [how to index large instances efficiently](#how-to-index-large-instances-efficiently) below.
| `Elasticsearch indexing` | Enables or disables Elasticsearch indexing and creates an empty index if one does not already exist. You may want to enable indexing but disable search 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 the background indexer which tracks data changes and ensures new data is indexed. |
| `Pause Elasticsearch indexing` | Enables or disables temporary indexing pause. This is useful for cluster migration/reindexing. All changes are still tracked, but they are not committed to the Elasticsearch index until resumed. |
| `Search with Elasticsearch enabled` | Enables or disables using Elasticsearch in search. |
| `URL` | The URL of your Elasticsearch instance. Use a comma-separated list to support clustering (for example, `http://host1, https://host2:9200`). If your Elasticsearch instance is password-protected, use the `Username` and `Password` fields described below. Alternatively, use inline credentials such as `http://<username>:<password>@<elastic_host>:9200/`. |
| `Username` | The `username` of your Elasticsearch instance. |
| `Password` | The password of your Elasticsearch instance. |
| `Number of Elasticsearch shards` | Elasticsearch indices are split into multiple shards for performance reasons. In general, you should use at least 5 shards, and indices with tens of millions of documents need to have more shards ([see below](#guidance-on-choosing-optimal-cluster-configuration)). Changes to this value do not take effect until the index is recreated. You can read more about tradeoffs in the [Elasticsearch documentation](https://www.elastic.co/guide/en/elasticsearch/reference/current/scalability.html). |
| `Number of Elasticsearch replicas` | Each Elasticsearch shard can have a number of replicas. These are a complete copy of the shard, and can provide increased query performance or resilience against hardware failure. Increasing this value increases total disk space required by the index. |
| `Limit the number of namespaces and projects that can be indexed` | Enabling this allows you to select namespaces and projects to index. All other namespaces and projects use database search instead. If you enable this option but do not select any namespaces or projects, none are indexed. [Read more below](#limit-the-number-of-namespaces-and-projects-that-can-be-indexed).|
| `Using AWS OpenSearch Service with IAM credentials` | Sign your OpenSearch requests using [AWS IAM authorization](https://docs.aws.amazon.com/IAM/latest/UserGuide/id_credentials_access-keys.html), [AWS EC2 Instance Profile Credentials](https://docs.aws.amazon.com/codedeploy/latest/userguide/getting-started-create-iam-instance-profile.html#getting-started-create-iam-instance-profile-cli), or [AWS ECS Tasks Credentials](https://docs.aws.amazon.com/AmazonECS/latest/userguide/task-iam-roles.html). Refer to [Identity and Access Management in Amazon OpenSearch Service](https://docs.aws.amazon.com/opensearch-service/latest/developerguide/ac.html) for details of AWS hosted OpenSearch domain access policy configuration. |
| `Maximum bulk request size (MiB)` | Used by the GitLab Ruby and Golang-based indexer processes. This setting indicates how much data must be collected (and stored in memory) in a given indexing process before submitting the payload to the Elasticsearch Bulk API. For the GitLab Golang-based indexer, you should use this setting with `Bulk request concurrency`. `Maximum bulk request size (MiB)` must accommodate the resource constraints of both the Elasticsearch hosts and the hosts running the GitLab Golang-based indexer from either the `gitlab-rake` command or the Sidekiq tasks. |
| `Bulk request concurrency` | The Bulk request concurrency indicates how many of the GitLab Golang-based indexer processes (or threads) can run in parallel to collect data to subsequently submit to the Elasticsearch Bulk API. This increases indexing performance, but fills the Elasticsearch bulk requests queue faster. This setting should be used together with the Maximum bulk request size setting (see above) and needs to accommodate the resource constraints of both the Elasticsearch hosts and the hosts running the GitLab Golang-based indexer either from the `gitlab-rake` command or the Sidekiq tasks. |
| `Client request timeout` | Elasticsearch HTTP client request timeout value in seconds. `0` means using the system default timeout value, which depends on the libraries that GitLab application is built upon. |
WARNING:
Increasing the values of `Maximum bulk request size (MiB)` and `Bulk request concurrency` can negatively impact
Sidekiq performance. Return them to their default values if you see increased `scheduling_latency_s` durations
To access and perform operations in Elasticsearch, GitLab requires a role with the following privileges:
```json
{
"cluster": ["monitor"],
"indices": [
{
"names": ["gitlab-*"],
"privileges": [
"create_index",
"delete_index",
"view_index_metadata",
"read",
"manage",
"write"
]
}
]
}
```
For more information, see [Elasticsearch security privileges](https://www.elastic.co/guide/en/elasticsearch/reference/current/security-privileges.html).
#### AWS OpenSearch Service with fine-grained access control
For more information, see [Identity and Access Management in Amazon OpenSearch Service](https://docs.aws.amazon.com/opensearch-service/latest/developerguide/ac.html).
##### Connecting with a master user in the internal database
When using fine-grained access control with a user in the internal database, you should use HTTP basic
authentication to connect to OpenSearch. You can provide the master username and password as part of the
OpenSearch URL or in the **Username** and **Password** text boxes in the Advanced Search settings. See
[Tutorial: Internal user database and HTTP basic authentication](https://docs.aws.amazon.com/opensearch-service/latest/developerguide/fgac-walkthrough-basic.html) for details.
When using fine-grained access control with IAM credentials, you can provide the credentials in the **AWS OpenSearch IAM credentials** section in the Advanced Search settings.
The following permissions are required for Advanced Search. See [Creating roles](https://docs.aws.amazon.com/opensearch-service/latest/developerguide/fgac.html#fgac-roles) for details.
```json
{
"cluster_permissions": [
"cluster_composite_ops",
"cluster_monitor"
],
"index_permissions": [
{
"index_patterns": [
"gitlab*"
],
"allowed_actions": [
"data_access",
"manage_aliases",
"search",
"create_index",
"delete",
"manage"
]
},
{
"index_patterns": [
"*"
],
"allowed_actions": [
"indices:admin/aliases/get",
"indices:monitor/stats"
]
}
]
}
```
The index pattern `*` requires a few permissions for Advanced Search to work.
any subgroups and projects belonging to those subgroups to be indexed as well.
Advanced Search only provides cross-group code/commit search (global) if all name-spaces are indexed. In this particular scenario where only a subset of namespaces are indexed, a global search does not provide a code or commit scope. This is possible only in the scope of an indexed namespace. There is no way to code/commit search in multiple indexed namespaces (when only a subset of namespaces has been indexed). For example if two groups are indexed, there is no way to run a single code search on both. You can only run a code search on the first group and then on the second.
If no namespaces or projects are selected, no Advanced Search indexing takes place.
WARNING:
If you have already indexed your instance, you must regenerate the index to delete all existing data
for filtering to work correctly. To do this, run the Rake tasks `gitlab:elastic:recreate_index` and
`gitlab:elastic:clear_index_status`. Afterwards, removing a namespace or a project from the list deletes the data
from the Elasticsearch index as expected.
## Enable custom language analyzers
You can improve the language support for Chinese and Japanese languages by utilizing [`smartcn`](https://www.elastic.co/guide/en/elasticsearch/plugins/current/analysis-smartcn.html) and/or [`kuromoji`](https://www.elastic.co/guide/en/elasticsearch/plugins/current/analysis-kuromoji.html) analysis plugins from Elastic.
1. Install the desired plugins, refer to [Elasticsearch documentation](https://www.elastic.co/guide/en/elasticsearch/plugins/7.9/installation.html) for plugins installation instructions. The plugins must be installed on every node in the cluster, and each node must be restarted after installation. For a list of plugins, see the table later in this section.
| `Enable Chinese (smartcn) custom analyzer: Indexing` | Enables or disables Chinese language support using [`smartcn`](https://www.elastic.co/guide/en/elasticsearch/plugins/current/analysis-smartcn.html) custom analyzer for newly created indices.|
| `Enable Chinese (smartcn) custom analyzer: Search` | Enables or disables using [`smartcn`](https://www.elastic.co/guide/en/elasticsearch/plugins/current/analysis-smartcn.html) fields for Advanced Search. Only enable this after [installing the plugin](https://www.elastic.co/guide/en/elasticsearch/plugins/current/analysis-smartcn.html), enabling custom analyzer indexing and recreating the index.|
| `Enable Japanese (kuromoji) custom analyzer: Indexing` | Enables or disables Japanese language support using [`kuromoji`](https://www.elastic.co/guide/en/elasticsearch/plugins/current/analysis-kuromoji.html) custom analyzer for newly created indices.|
| `Enable Japanese (kuromoji) custom analyzer: Search` | Enables or disables using [`kuromoji`](https://www.elastic.co/guide/en/elasticsearch/plugins/current/analysis-kuromoji.html) fields for Advanced Search. Only enable this after [installing the plugin](https://www.elastic.co/guide/en/elasticsearch/plugins/current/analysis-kuromoji.html), enabling custom analyzer indexing and recreating the index.|
1. On the left sidebar, select **Settings > Advanced Search**.
1. Expand **Advanced Search**.
1. Clear the **Pause Elasticsearch indexing** checkbox.
## Zero downtime reindexing
The idea behind this reindexing method is to leverage the [Elasticsearch reindex API](https://www.elastic.co/guide/en/elasticsearch/reference/current/docs-reindex.html)
and Elasticsearch index alias feature to perform the operation. We set up an index alias which connects to a
`primary` index which is used by GitLab for reads/writes. When reindexing process starts, we temporarily pause
the writes to the `primary` index. Then, we create another index and invoke the Reindex API which migrates the
index data onto the new index. After the reindexing job is complete, we switch to the new index by connecting the
### Trigger the reindex via the Advanced Search administration
> - [Introduced](https://gitlab.com/gitlab-org/gitlab/-/merge_requests/34069) in GitLab 13.2.
> - A scheduled index deletion and the ability to cancel it was [introduced](https://gitlab.com/gitlab-org/gitlab/-/merge_requests/38914) in GitLab 13.3.
> - Support for retries during reindexing was [introduced](https://gitlab.com/gitlab-org/gitlab/-/merge_requests/55681) in GitLab 13.12.
To trigger the reindexing process:
1. Sign in to your GitLab instance as an administrator.
The slice multiplier calculates the [number of slices during reindexing](https://www.elastic.co/guide/en/elasticsearch/reference/current/docs-reindex.html#docs-reindex-slice).
| [`sudo gitlab-rake gitlab:elastic:info`](https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/lib/tasks/gitlab/elastic.rake) | Outputs debugging information for the Advanced Search integration. |
| [`sudo gitlab-rake gitlab:elastic:index_projects`](https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/lib/tasks/gitlab/elastic.rake) | Iterates over all projects, and queues Sidekiq jobs to index them in the background. It can only be used after the index is created. |
| [`sudo gitlab-rake gitlab:elastic:index_projects_status`](https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/lib/tasks/gitlab/elastic.rake) | Determines the overall status of the indexing. It is done by counting the total number of indexed projects, dividing by a count of the total number of projects, then multiplying by 100. |
| [`sudo gitlab-rake gitlab:elastic:clear_index_status`](https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/lib/tasks/gitlab/elastic.rake) | Deletes all instances of IndexStatus for all projects. This command results in a complete wipe of the index, and it should be used with caution. |
| [`sudo gitlab-rake gitlab:elastic:create_empty_index`](https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/lib/tasks/gitlab/elastic.rake) | Generates empty indices (the default index and a separate issues index) and assigns an alias for each on the Elasticsearch side only if it doesn't already exist. |
| [`sudo gitlab-rake gitlab:elastic:delete_index`](https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/lib/tasks/gitlab/elastic.rake) | Removes the GitLab indices and aliases (if they exist) on the Elasticsearch instance. |
| [`sudo gitlab-rake gitlab:elastic:recreate_index`](https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/lib/tasks/gitlab/elastic.rake) | Wrapper task for `gitlab:elastic:delete_index` and `gitlab:elastic:create_empty_index`. |
| [`sudo gitlab-rake gitlab:elastic:index_snippets`](https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/lib/tasks/gitlab/elastic.rake) | Performs an Elasticsearch import that indexes the snippets data. |
| [`sudo gitlab-rake gitlab:elastic:projects_not_indexed`](https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/lib/tasks/gitlab/elastic.rake) | Displays which projects are not indexed. |
| [`sudo gitlab-rake gitlab:elastic:reindex_cluster`](https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/lib/tasks/gitlab/elastic.rake) | Schedules a zero-downtime cluster reindexing task. This feature should be used with an index that was created after GitLab 13.0. |
| [`sudo gitlab-rake gitlab:elastic:mark_reindex_failed`](https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/lib/tasks/gitlab/elastic.rake) | Mark the most recent re-index job as failed. |
| [`sudo gitlab-rake gitlab:elastic:list_pending_migrations`](https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/lib/tasks/gitlab/elastic.rake) | List pending migrations. Pending migrations include those that have not yet started, have started but not finished, and those that are halted. |
| [`sudo gitlab-rake gitlab:elastic:estimate_cluster_size`](https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/lib/tasks/gitlab/elastic.rake) | Get an estimate of cluster size based on the total repository size. |
### Guidance on choosing optimal cluster configuration
For basic guidance on choosing a cluster configuration you may refer to [Elastic Cloud Calculator](https://cloud.elastic.co/pricing). You can find more information below.
- Generally, you want to use at least a 2-node cluster configuration with one replica, which allows you to have resilience. If your storage usage is growing quickly, you may want to plan horizontal scaling (adding more nodes) beforehand.
- It's not recommended to use HDD storage with the search cluster, because it takes a hit on performance. It's better to use SSD storage (NVMe or SATA SSD drives for example).
- You can use the [GitLab Performance Tool](https://gitlab.com/gitlab-org/quality/performance) to benchmark search performance with different search cluster sizes and configurations.
-`Heap size` should be set to no more than 50% of your physical RAM. Additionally, it shouldn't be set to more than the threshold for zero-based compressed oops. The exact threshold varies, but 26 GB is safe on most systems, but can also be as large as 30 GB on some systems. See [Heap size settings](https://www.elastic.co/guide/en/elasticsearch/reference/current/important-settings.html#heap-size-settings) and [Setting JVM options](https://www.elastic.co/guide/en/elasticsearch/reference/current/jvm-options.html) for more details.
- Number of CPUs (CPU cores) per node usually corresponds to the `Number of Elasticsearch shards` setting described below.
- A good guideline is to ensure you keep the number of shards per node below 20 per GB heap it has configured. A node with a 30 GB heap should therefore have a maximum of 600 shards, but the further below this limit you can keep it the better. This generally helps the cluster stay in good health.
- Another consideration is the number of documents. To determine the number of shards to use, sum the numbers in the **Main menu > Admin > Dashboard > Statistics** pane (the number of documents to be indexed), divide by 5 million, and add 5. For example:
-`refresh_interval` is a per index setting. You may want to adjust that from default `1s` to a bigger value if you don't need data in real-time. This changes how soon you see fresh results. If that's important for you, you should leave it as close as possible to the default value.
- You might want to raise [`indices.memory.index_buffer_size`](https://www.elastic.co/guide/en/elasticsearch/reference/current/indexing-buffer.html) to 30% or 40% if you have a lot of heavy indexing operations.
### Advanced Search integration settings guidance
- The `Number of Elasticsearch shards` setting usually corresponds with the number of CPUs available in your cluster. For example, if you have a 3-node cluster with 4 cores each, this means you benefit from having at least 3*4=12 shards in the cluster. It's only possible to change the shards number by using [Split index API](https://www.elastic.co/guide/en/elasticsearch/reference/current/indices-split-index.html) or by reindexing to a different index with a changed number of shards.
- The `Number of Elasticsearch replicas` setting should most of the time be equal to `1` (each shard has 1 replica). Using `0` is not recommended, because losing one node corrupts the index.
### How to index large instances efficiently
This section may be helpful in the event that the other
[basic instructions](#enable-advanced-search) cause problems
due to large volumes of data being indexed.
WARNING:
Indexing a large instance generates a lot of Sidekiq jobs.
1. Indexing large Git repositories can take a while. To speed up the process, you can [tune for indexing speed](https://www.elastic.co/guide/en/elasticsearch/reference/current/tune-for-indexing-speed.html#tune-for-indexing-speed):
- You can temporarily disable [`refresh`](https://www.elastic.co/guide/en/elasticsearch/reference/current/indices-refresh.html), the operation responsible for making changes to an index available to search.
- You can set the number of replicas to 0. This setting controls the number of copies each primary shard of an index has. Thus, having 0 replicas effectively disables the replication of shards across nodes, which should increase the indexing performance. This is an important trade-off in terms of reliability and query performance. It is important to remember to set the replicas to a considered value after the initial indexing is complete.
In our experience, you can expect a 20% decrease in indexing time. After completing indexing in a later step, you can return `refresh` and `number_of_replicas` to their desired settings.
NOTE:
This step is optional but may help significantly speed up large indexing operations.
```shell
curl --request PUT localhost:9200/gitlab-production/_settings --header 'Content-Type: application/json' \
Whenever a change or deletion is made to an indexed GitLab object (a merge request description is changed, a file is deleted from the default 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 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 letting Elasticsearch merge and reclaim space automatically, with the default settings. From [Lucene's Handling of Deleted Documents](https://www.elastic.co/blog/lucenes-handling-of-deleted-documents "Lucene's Handling of Deleted Documents"), _"Overall, besides perhaps decreasing the maximum segment size, it is best to leave Lucene defaults as-is and not fret too much about when deletes are reclaimed."_
- 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 allows merging to happen more frequently.
curl --request PUT localhost:9200/gitlab-production/_settings ---header 'Content-Type: application/json' \
--data '{
"index" : {
"merge.policy.max_merged_segment": "2gb"
}
}'
```
- You can also adjust `index.merge.policy.reclaim_deletes_weight`, which controls how aggressively deletions are targeted. But this can lead to costly merge decisions, so we recommend not changing this unless you understand the tradeoffs.
```shell
curl --request PUT localhost:9200/gitlab-production/_settings ---header 'Content-Type: application/json' \
--data '{
"index" : {
"merge.policy.reclaim_deletes_weight": "3.0"
}
}'
```
- Do not do a [force merge](https://www.elastic.co/guide/en/elasticsearch/reference/current/indices-forcemerge.html "Force Merge") to remove deleted documents. A warning in the [documentation](https://www.elastic.co/guide/en/elasticsearch/reference/current/indices-forcemerge.html "Force Merge") states that this can lead to very large segments that may never get reclaimed, and can also cause significant performance or availability issues.
## Index large instances with dedicated Sidekiq nodes or processes
Indexing a large instance can be a lengthy and resource-intensive process that has the potential
of overwhelming Sidekiq nodes and processes. This negatively affects the GitLab performance and
availability.
As GitLab allows you to start multiple Sidekiq processes, you can create an
additional process dedicated to indexing a set of queues (or queue group). This way, you can
ensure that indexing queues always have a dedicated worker, while the rest of the queues have
For this purpose, use the [queue selectors](../../administration/sidekiq/processing_specific_job_classes.md#queue-selectors)
option that allows a more general selection of queue groups using a [worker matching query](../../administration/sidekiq/processing_specific_job_classes.md#worker-matching-query).
