Common Errors and Solutions

Warning:
GA releases for CockroachDB v23.1 are no longer supported. Cockroach Labs will stop providing LTS Assistance Support for v23.1 LTS releases on November 13, 2025. Prior to that date, upgrade to a more recent version to continue receiving support. For more details, refer to the Release Support Policy.

This page helps you understand and resolve error messages written to stderr or your logs.

Topic Message
Client connection connection refused
Client connection node is running secure mode, SSL connection required
Transaction retries restart transaction
Node startup node belongs to cluster <cluster ID> but is attempting to connect to a gossip network for cluster <another cluster ID>
Node configuration clock synchronization error: this node is more than 500ms away from at least half of the known nodes
Node configuration open file descriptor limit of <number> is under the minimum required <number>
Replication replicas failing with "0 of 1 store with an attribute matching []; likely not enough nodes in cluster"
Split failed split failed while applying backpressure; are rows updated in a tight loop?
Deadline exceeded context deadline exceeded
Incremental backups protected ts verification error...
Ambiguous results result is ambiguous
Import key collision checking for key collisions: ingested key collides with an existing one
SQL memory budget exceeded memory budget exceeded

connection refused

This message indicates a client is trying to connect to a node that is either not running or is not listening on the specified interfaces (i.e., hostname or port).

To resolve this issue, do one of the following:

If you're not sure what the IP address/hostname and port values might have been, you can look in the node's logs.

If necessary, you can also shut down and then restart the node:

icon/buttons/copy
$ cockroach start [flags]

node is running secure mode, SSL connection required

This message indicates that the cluster is using TLS encryption to protect network communication, and the client is trying to open a connection without using the required TLS certificates.

To resolve this issue, use the cockroach cert create-client command to generate a client certificate and key for the user trying to connect. For a secure deployment tutorial, including generating security certificates and connecting clients, see Manual Deployment.

restart transaction

Messages with the error code 40001 and the string restart transaction are known as transaction retry errors. These indicate that a transaction failed due to contention with another concurrent or recent transaction attempting to write to the same data. The transaction needs to be retried by the client.

In most cases, the correct actions to take when encountering transaction retry errors are:

  1. Update your application to support client-side retry handling when transaction retry errors are encountered. Follow the guidance for the specific error type.

  2. Take steps to minimize transaction retry errors in the first place. This means reducing transaction contention overall, and increasing the likelihood that CockroachDB can automatically retry a failed transaction.

node belongs to cluster <cluster ID> but is attempting to connect to a gossip network for cluster <another cluster ID>

This message usually indicates that a node tried to connect to a cluster, but the node is already a member of a different cluster. This is determined by metadata in the node's data directory. To resolve this issue, do one of the following:

  • Choose a different directory to store the CockroachDB data:

    icon/buttons/copy
    $ cockroach start [flags] --store=[new directory] --join=[cluster host]:26257
    
  • Remove the existing directory and start a node joining the cluster again:

    icon/buttons/copy
    $ rm -r cockroach-data/
    
    icon/buttons/copy
    $ cockroach start [flags] --join=[cluster host]:26257
    

clock synchronization error: this node is more than 500ms away from at least half of the known nodes

This error indicates that a node has spontaneously shut down because it detected that its clock is out of sync with at least half of the other nodes in the cluster by 80% of the maximum offset allowed (500ms by default). CockroachDB requires moderate levels of clock synchronization to preserve data consistency, so the node shutting down in this way avoids the risk of consistency anomalies.

To prevent this from happening, you should run clock synchronization software on each node. For guidance on synchronizing clocks, see the tutorial for your deployment environment:

Environment Recommended Approach
Manual Use NTP with Google's external NTP service.
AWS Use the Amazon Time Sync Service.
Azure Disable Hyper-V time synchronization and use NTP with Google's external NTP service.
Digital Ocean Use NTP with Google's external NTP service.
GCE Use NTP with Google's internal NTP service.

open file descriptor limit of <number> is under the minimum required <number>

CockroachDB can use a large number of open file descriptors, often more than is available by default. This message indicates that the machine on which a CockroachDB node is running is under CockroachDB's recommended limits.

For more details on CockroachDB's file descriptor limits and instructions on increasing the limit on various platforms, see File Descriptors Limit.

replicas failing with "0 of 1 store with an attribute matching []; likely not enough nodes in cluster

When running a single-node cluster

When running a single-node CockroachDB cluster, an error about replicas failing will eventually show up in the node's log files, for example:

E160407 09:53:50.337328 storage/queue.go:511  [replicate] 7 replicas failing with "0 of 1 store with an attribute matching []; likely not enough nodes in cluster"

This happens because CockroachDB expects three nodes by default. If you do not intend to add additional nodes, you can stop this error by using ALTER RANGE ... CONFIGURE ZONE to update your default zone configuration to expect only one node:

icon/buttons/copy
# Insecure cluster:
$ cockroach sql --execute="ALTER RANGE default CONFIGURE ZONE USING num_replicas=1;" --insecure
icon/buttons/copy
# Secure cluster:
$ cockroach sql --execute="ALTER RANGE default CONFIGURE ZONE USING num_replicas=1;" --certs-dir=[path to certs directory]

The zone's replica count is reduced to 1. For more information, see ALTER RANGE ... CONFIGURE ZONE and Replication Controls.

When running a multi-node cluster

When running a multi-node CockroachDB cluster, if you see an error like the one above about replicas failing, some nodes might not be able to talk to each other. For recommended actions, see Cluster Setup Troubleshooting.

split failed while applying backpressure; are rows updated in a tight loop?

In CockroachDB, a table row is stored on disk as a key-value pair. Whenever the row is updated, CockroachDB also stores a distinct version of the key-value pair to enable concurrent request processing while guaranteeing consistency (see multi-version concurrency control (MVCC)). All versions of a key-value pair belong to a larger "range" of the total key space, and the historical versions remain until the garbage collection period defined by the gc.ttlseconds variable in the applicable zone configuration has passed. Once a range reaches a size threshold, CockroachDB splits the range into two ranges. However, this message indicates that a range cannot be split as intended.

One possible cause is that the range consists only of MVCC version data due to a row being repeatedly updated, and the range cannot be split because doing so would spread MVCC versions for a single row across multiple ranges.

To resolve this issue, make sure you are not repeatedly updating a single row. If frequent updates of a row are necessary, consider one of the following:

  • Reduce the gc.ttlseconds variable in the applicable zone configuration to reduce the garbage collection period and prevent such a large build-up of historical values.
  • If a row contains large columns that are not being updated with other columns, put the large columns in separate column families.

context deadline exceeded

This message occurs when a component of CockroachDB gives up because it was relying on another component that has not behaved as expected, for example, another node dropped a network connection. To investigate further, look in the node's logs for the primary failure that is the root cause.

protected ts verification error

Messages that begin with protected ts verification error… indicate that your incremental backup failed because the data you are trying to backup was garbage collected. This happens when incremental backups are taken less frequently than the garbage collection periods for any of the objects in the base backup. For example, if your incremental backups recur daily, but the garbage collection period of one table in your backup is less than one day, all of your incremental backups will fail.

The error message will specify which part of your backup is causing the failure. For example, range span: /Table/771 indicates that table 771 is part of the problem. You can then inspect this table by running SELECT * FROM crdb_internal.tables WHERE id=771. You can also run SHOW ZONE CONFIGURATIONS and look for any gc.ttlseconds values that are set lower than your incremental backup frequency.

To resolve this issue, take a new full backup after doing either of the following:

Also, consider using scheduled backups that use protected timestamps to ensure that the data to be backed up is protected from garbage collection until it has been successfully backed up. This active management of protected timestamps means that you can run scheduled backups at a cadence independent from the GC TTL of the data. For more detail, see Protected timestamps and scheduled backups.

result is ambiguous

In a distributed system, some errors can have ambiguous results. For example, if you receive a connection closed error while processing a COMMIT statement, you cannot tell whether the transaction successfully committed or not. These errors are possible in any database, but CockroachDB is somewhat more likely to produce them than other databases because ambiguous results can be caused by failures between the nodes of a cluster. These errors are reported with the PostgreSQL error code 40003 (statement_completion_unknown) and the message result is ambiguous.

Ambiguous errors can be caused by nodes crashing, network failures, or timeouts. If you experience a lot of these errors when things are otherwise stable, look for performance issues. Note that ambiguity is only possible for the last statement of a transaction (COMMIT or RELEASE SAVEPOINT) or for statements outside a transaction. If a connection drops during a transaction that has not yet tried to commit, the transaction will definitely be aborted.

In general, you should handle ambiguous errors the same way as connection closed errors. If your transaction is idempotent, it is safe to retry it on ambiguous errors. UPSERT operations are typically idempotent, and other transactions can be written to be idempotent by verifying the expected state before performing any writes. Increment operations such as UPDATE my_table SET x=x+1 WHERE id=$1 are typical examples of operations that cannot easily be made idempotent. If your transaction is not idempotent, then you should decide whether to retry or not based on whether it would be better for your application to apply the transaction twice or return an error to the user.

checking for key collisions: ingested key collides with an existing one

When importing into an existing table with IMPORT INTO, this error occurs because the rows in the import file conflict with an existing primary key or another UNIQUE constraint on the table. The import will fail as a result. IMPORT INTO is an insert-only statement, so you cannot use it to update existing rows. To update rows in an existing table, use the UPDATE statement.

memory budget exceeded

If a node runs out of its allocated SQL memory (the memory allocated to the SQL layer), a memory budget exceeded error occurs.

To mitigate this issue, ensure that the node has enough RAM and that enough memory is allocated to the SQL layer. The best approach depends upon the cluster's workload. Try the following approaches:

  • If you find queries that are consuming too much memory, cancel the queries to free up memory usage. For information on optimizing query performance, see SQL Performance Best Practices.

  • Increase the amount of memory on the node. Cockroach Labs recommends that you use the same hardware, operating system, and software configuration on each node.

  • Increase --max-sql-memory on the node. A memory budget exceeded error is an early warning that the cockroach process on a node is at risk of crashing due to an out-of-memory (OOM) crash. To protect the node, CockroachDB fails the query.

However, do not set --max-sql-memory too high. The operating system dynamically increases the amount of memory available to the cockroach process, and by default, 25% of the memory allocated to the cockroach process is reserved for the SQL layer. If the demand exceeds the amount of RAM on the node, the cockroach process may crash or become very slow by falling back to using disk-based swap. Try different values and monitor your cluster's performance. Avoid increasing the value further as soon the total memory usage under load grows beyond 80% of overall capacity available to the process.

  • For disk-spilling operations such as hash joins that are memory-intensive, consider allocating more memory to the operation before it spills to disk and risks consuming more memory. To do this, increase the value of the sql.distsql.temp_storage.workmem cluster setting. This improves the performance of the query, with the risk of a reduction in the concurrency of the workload. Try different values and monitor your cluster's performance.

For example, if a query contains a hash join that requires 128 MiB of memory before spilling to disk, you can set sql.distsql.temp_storage.workmem=64MiB and --max-sql-memory=1GiB to allow the query to run 16 times concurrently. A 17th concurrent instance of the query exceeds --max-sql-memory and produces a memory budget exceeded error. To allow only 8 instances to run in parallel but allow all queries to finish without spilling to disk, set sql.distsql.temp_storage.workmem to 128MiB.

For more information, refer to:

Something else?

Try searching the rest of our docs for answers or using our other support resources, including:


Yes No