doc/administration/settings/import_and_export_settings.md
{{< details >}}
{{< /details >}}
Settings for import- and export-related features.
Before you can import projects from other systems, you must enable the import source for that system.
Only import projects from sources you trust. If you import a project from an untrusted source,
an attacker could steal your sensitive data. For example, an imported project
with a malicious .gitlab-ci.yml file could allow an attacker to exfiltrate group CI/CD variables.
GitLab Self-Managed administrators can reduce their attack surface by disabling import sources they don't need:
To enable the export of projects and their data:
{{< history >}}
{{< /history >}}
Migration of groups and projects by direct transfer is disabled by default. To enable migration of groups and projects by direct transfer:
The same setting
is available in the API as the
bulk_import_enabled attribute.
{{< history >}}
offline_transfer_exports. Disabled by default.{{< /history >}}
[!flag] The availability of this feature is controlled by a feature flag. For more information, see the history.
Prerequisites:
By default, offline transfer supports only AWS S3 and Google Cloud Storage. Turn on this setting to also allow S3-compatible providers, such as MinIO.
[!warning] When you enable this setting, users who can perform offline transfers can supply an arbitrary
endpointURL in the offline transfer object storage configuration. GitLab then sends requests to that endpoint. Enable this setting only if you trust the users who can perform offline transfers.
To allow S3-compatible object storage for offline transfer:
{{< history >}}
offline_transfer_exports, offline_transfer_imports, and offline_transfer_ui. Disabled by default.{{< /history >}}
[!flag] The availability of this feature is controlled by a feature flag. For more information, see the history.
Turn on this setting to allow offline transfer to authenticate with Google Cloud Storage by using Application Default Credentials (ADC).
With every other object storage provider, the user who creates an export or an import supplies the
credentials. With ADC, no user supplies credentials. GitLab stores only the Google Cloud project ID
and resolves the credentials for each request from the environment of the instance, either from the
Compute Engine metadata server or from the GOOGLE_APPLICATION_CREDENTIALS environment variable.
Prerequisites:
[!warning] An offline transfer that uses ADC acts with every Cloud Storage permission that the service account of the instance holds. That service account is usually more privileged than any individual user, so a user who creates an ADC transfer can reach buckets they hold no credentials for.
To limit this risk, GitLab applies the following restrictions, which you cannot turn off:
Only administrators can use Application Default Credentials for offline transfer.gitlab-offline-transfer-. This prefix keeps ADC transfers away
from the buckets that the instance uses for its own object storage, such as uploads, job artifacts,
and LFS objects.The Google Cloud project ID that you provide for a transfer does not limit which buckets that
transfer can reach. Bucket names are globally unique in Cloud Storage, so an ADC transfer can use any
bucket that the service account can access and whose name starts with gitlab-offline-transfer-,
in any Google Cloud project.
GitLab checks these restrictions when a user creates a transfer. If you turn off this setting, users can no longer create ADC transfers, but transfers that already started continue to run.
To allow Application Default Credentials for offline transfer:
The same setting
is available in the API as the
allow_application_default_credentials_for_offline_transfer attribute.
{{< history >}}
export_audit_events. Disabled by default.export_audit_events removed.{{< /history >}}
Enable silent admin exports to prevent audit events when instance administrators trigger a project or group file export or download the export file. Exports from non-administrators still generate audit events.
To enable silent admin project and group file exports:
{{< history >}}
importer_user_mapping. Disabled by default.importer_user_mapping removed.{{< /history >}}
To allow mapping of imported user contributions to administrators:
{{< history >}}
importer_user_mapping_allow_bypass_of_confirmation. Disabled by default.importer_user_mapping_allow_bypass_of_confirmation removed.{{< /history >}}
Prerequisites:
To skip confirmation when administrators reassign placeholder users:
When this setting is enabled, administrators can reassign contributions and memberships to non-bot users with any of the following states:
activebannedblockedblocked_pending_approvaldeactivatedldap_blockedTo modify the maximum file size for exports in GitLab:
To modify the maximum file size for imports in GitLab:
This setting applies only to repositories imported from a GitLab export file.
If you choose a size larger than the configured value for the web server, you may receive errors. See the troubleshooting section for more details.
For GitLab.com repository size limits, read accounts and limit settings.
By default, the maximum remote file size for imports from external object storages (for example, AWS) is 10 GiB.
To modify this setting:
0 for no file size limit.By default, the maximum download file size for imports by direct transfer is 5 GiB.
To modify this setting:
0 for no file size limit.When you import a project using file exports or direct transfer, you can specify the maximum decompressed file size for imported archives. The default value is 25 GiB.
When you import a compressed file, the decompressed size cannot exceed the maximum decompressed file size limit. If the decompressed size exceeds the configured limit, the following error is returned:
Decompressed archive size validation failed.
To modify this setting:
When you import a project, you can specify the maximum time out for decompressing imported archives. The default value is 210 seconds.
To modify the maximum decompressed file size for imports in GitLab:
You can specify the maximum number of import jobs that are executed simultaneously for:
The job limit is not applied when importing merge requests because there is a hard-coded limit for merge requests to avoid overloading servers.
The default job limit is:
To modify this setting:
{{< history >}}
{{< /history >}}
Direct transfer exports can consume a significant amount of resources.
To prevent using up the database or Sidekiq processes,
administrators can configure the concurrent_relation_batch_export_limit setting.
The default value is 8 jobs, which corresponds to a
reference architecture for up to 40 RPS or 2,000 users.
If you encounter PG::QueryCanceled: ERROR: canceling statement due to statement timeout errors
or jobs getting interrupted due to Sidekiq memory limits, you might want to reduce this number.
If you have enough resources, you can increase this number to process more concurrent export jobs.
To modify this setting, send an API request to /api/v4/application/settings
with concurrent_relation_batch_export_limit.
For more information, see application settings API.
{{< history >}}
{{< /history >}}
To further manage memory usage and database load, use the relation_export_batch_size setting to control the number of records processed in each batch during export operations.
The default value is 50 records per batch. To modify this setting, send an API request to /api/v4/application/settings with relation_export_batch_size.
For more information, see application settings API.
Help page documentation base url is blocked: execution expiredWhile enabling application settings like import source, you might get a Help page documentation base url is blocked: execution expired
error. To work around this error:
docs.gitlab.com, or the redirect help documentation pages URL, to the
allowlist.