Before you begin reading this guide, we recommend you run Elasticsearch Error Check-Up which can resolve issues that cause many errors.
This guide will help you check for common problems that cause the log ” Recovery failed ” to appear. It’s important to understand the issues related to the log, so to get started, read the general overview on common issues and tips related to the Elasticsearch concepts: recovery.
Advanced users might want to skip right to the common problems section in each concept or try running the Check-Up which analyses ES to pinpoint the cause of many errors and provides suitable actionable recommendations how to resolve them (free tool that requires no installation).
Overview
In Elasticsearch, recovery refers to the process of recovering an index or shard when something goes wrong. There are many ways to recover an index or shard, such as by re-indexing the data from a backup / failover cluster to the current one, or by restoring from an Elasticsearch snapshot. Alternatively, Elasticsearch performs recoveries automatically, such as when a node restarts or disconnects and connects again. There is an API to check the updated status of index / shard recoveries.
GET //_recoveryGET /_recovery
In summary, recovery can happen in the following scenarios:
- Node startup or failure (local store recovery )
- Replication of primary shards to replica shards
- Relocation of a shard to a different node in the same cluster
- Restoring a snapshot
Examples
Getting recovery information about several indices:
GET my_index1 GET my_index2/_recovery
Notes and good things to know
- When a node is disconnected from the cluster, all of its shards go to an unassigned state. After a certain amount of time, the shards will be allocated somewhere else on other nodes. This setting determines the number of concurrent shards per node that will be recovered.
PUT _cluster/settings{"transient":{"cluster.routing.allocation.node_concurrent_recoveries":3}}
- You can also control when to start recovery after a node disconnects. This is useful if the node just restarts, for example, because you may not want to initiate any recovery for such transient events.
PUT _all/_settings{"settings":{"index.unassigned.node_left.delayed_timeout":"6m"}}
- Elasticsearch limits the speed that is allocated to recovery in order to avoid overloading the cluster. This setting can be updated to make the recovery faster or slower, depending on your requirements.
PUT _cluster/settings{"transient":{"indices.recovery.max_bytes_per_sec":"100mb"}}
Log Context
Log “Recovery failed” classname is GatewayService.java
We extracted the following from Elasticsearch source code for those seeking an in-depth context :
} else { if (recovered.compareAndSet(false; true)) { threadPool.generic().execute(new AbstractRunnable() { Override public void onFailure(Exception e) { logger.warn("Recovery failed"; e); // we reset `recovered` in the listener don't reset it here otherwise there might be a race // that resets it to false while a new recover is already running? recoveryListener.onFailure("state recovery failed: " + e.getMessage()); }
Run the Check-Up to get customized recommendations like this:

Heavy merges detected in specific nodes

Description
A large number of small shards can slow down searches and cause cluster instability. Some indices have shards that are too small…

Recommendations Based on your specific ES deployment you should…
Based on your specific ES deployment you should…
X-PUT curl -H [a customized code snippet to resolve the issue]