[[add-repository]]
== Troubleshooting broken repositories

There are several situations where the <<health-api, Health API>> might report an issue
regarding the integrity of snapshot repositories in the cluster. The following pages explain
the recommended actions for diagnosing corrupted, unknown, and invalid repositories:

- <<diagnosing-corrupted-repositories>>
- <<diagnosing-unknown-repositories>>
- <<diagnosing-invalid-repositories>>

[[diagnosing-corrupted-repositories]]
=== Diagnosing corrupted repositories

Multiple {es} deployments are writing to the same snapshot repository. {es} doesn't
support this configuration and only one cluster is allowed to write to the same
repository. See <<snapshot-repository-contents>> for potential side-effects of
corruption of the repository contents, which may not be resolved by the following
guide.
To remedy the situation mark the repository as read-only or remove it from all the
other deployments, and re-add (recreate) the repository in the current deployment:

include::{es-ref-dir}/tab-widgets/troubleshooting/snapshot/corrupt-repository-widget.asciidoc[]


[[diagnosing-unknown-repositories]]
=== Diagnosing unknown repositories

When a snapshot repository is marked as "unknown", it means that an {es} node is
unable to instantiate the repository due to an unknown repository type. This is
usually caused by a missing plugin on the node. Make sure each node in the cluster
has the required plugins by following the following steps:

1. Retrieve the affected nodes from the affected resources section of the health report.
2. Use the <<cluster-nodes-info,nodes info API>> to retrieve the plugins installed on each node.
3. Cross reference this with a node that works correctly to find out which plugins are missing
and install the missing plugins.


[[diagnosing-invalid-repositories]]
=== Diagnosing invalid repositories

When an {es} node faces an unexpected exception when trying to instantiate a snapshot
repository, it will mark the repository as "invalid" and write a warning to the log file.
Use the following steps to diagnose the underlying cause of this issue:

1. Retrieve the affected nodes from the affected resources section of the health report.
2. Refer to the logs of the affected node(s) and search for the repository name.
You should be able to find logs that will contain relevant exception.
3. Try to resolve the errors reported.