Known Limitations in CockroachDB v23.2

On this page Carat arrow pointing down

This page describes newly identified limitations in the CockroachDB v23.2.6 release as well as unresolved limitations identified in earlier releases.

New limitations

PL/pgSQL Limitations

Support for PL/pgSQL features

Type Handling and Variable Declarations

Cursor Functionality

Exception Handling

Limitations in User-Defined Functions (UDFs) and Stored Procedures

SQL Optimizer and Read Committed Isolation

Optimizer and Locking Behavior

  • The SQL optimizer has limitations under certain isolation levels:
    • The new implementation of SELECT FOR UPDATE is not yet the default setting under SERIALIZABLE isolation. It can be used under SERIALIZABLE isolation by setting the optimizer_use_lock_op_for_serializable session setting to true. Tracking GitHub issue
    • SELECT FOR UPDATE does not lock completely-NULL column families in multi-column-family tables. Tracking GitHub issue

Read Committed Isolation Limitations

CAST expressions containing a subquery with an ENUM target are not supported

Physical cluster replication

  • Physical cluster replication is supported only on CockroachDB Self-Hosted in new v23.2 clusters. Physical Cluster Replication cannot be enabled on clusters that have been upgraded from a previous version of CockroachDB.
  • Cockroach Labs supports physical cluster replication up to the following scale:
    • Initial data load: 2TB
    • Read maximum: 1000 reads per second
    • Write maximum: 850 writes per second
  • Read queries are not supported on the standby cluster before cutover.
  • The primary and standby cluster cannot have different region topology. For example, replicating a multi-region primary cluster to a single-region standby cluster is not supported. Mismatching regions between a multi-region primary and standby cluster is also not supported.
  • Cutting back to the primary cluster after a cutover is a manual process. Refer to Cut back to the primary cluster. In addition, after cutover, to continue using physical cluster replication, you must configure it again.
  • Before cutover to the standby, the standby cluster does not support running backups or changefeeds.
  • After a cutover, there is no mechanism to stop applications from connecting to the original primary cluster. It is necessary to redirect application traffic manually, such as by using a network load balancer or adjusting DNS records.
  • Large data imports, such as those produced by RESTORE or IMPORT, may dramatically increase replication lag.

  • After the cutover process for physical cluster replication, scheduled changefeeds will continue on the promoted cluster. You will need to manage pausing or canceling the schedule on the promoted standby cluster to avoid two clusters running the same changefeed to one sink. Tracking GitHub issue

Unresolved limitations

Limitations for user-defined functions (UDFs)

Limitations on use of UDFs

User-defined functions are not currently supported in:

Limitations on expressions allowed within UDFs

The following are not currently allowed within the body of a UDF:

Table-level restore will not restore user-defined functions

RESTORE will not restore a table that references a UDF, unless you skip restoring the function with the skip_missing_udfs option. Alternatively, take a database-level backup to include everything needed to restore the table. Tracking GitHub Issue

Incorrect query plans for partitions with NULL values

In cases where the partition definition includes a comparison with NULL and a query constraint, incorrect query plans are returned. However, this case uses non-standard partitioning which defines partitions which could never hold values, so it is not likely to occur in production environments.

Tracking GitHub issue

null_ordered_last does not produce correct results with tuples

By default, CockroachDB orders NULLs before all other values. For compatibility with PostgreSQL, the null_ordered_last session variable was added, which changes the default to order NULLs after all other values. This works in most cases, due to some transformations CockroachDB makes in the optimizer to add extra ordering columns. However, it is broken when the ordering column is a tuple.

Tracking GitHub issue

Limitations for DROP OWNED BY

DROP OWNED BY drops all owned objects as well as any grants on objects not owned by the role.

DROP OWNED BY is not supported where role has system-level privileges

  • If the role for which you are trying to DROP OWNED BY was granted a system-level privilege (i.e., using the GRANT SYSTEM ... statement), the error shown below will be signalled. The workaround is to use SHOW SYSTEM GRANTS FOR {role} and then use REVOKE SYSTEM ... for each privilege in the result. For more information about this known limitation, see cockroachdb/cockroach#88149.

    ERROR: cannot perform drop owned by if role has synthetic privileges; foo has entries in system.privileges
    SQLSTATE: 0A000
    HINT: perform REVOKE SYSTEM ... for the relevant privileges foo has in system.privileges
    

Note that the phrase "synthetic privileges" in the above error message refers to system-level privileges.

Spatial features disabled for ARM Macs

Spatial features are disabled due to an issue with macOS code signing for the GEOS libraries. Users needing spatial features on an ARM Mac may instead use Rosetta to run the Intel binary or use the Docker image distribution. This is expected to be resolved in an upcoming 22.2 patch release.

GitHub tracking issue

Limited SQL cursor support

CockroachDB implements SQL cursor support with the following limitations:

  • DECLARE only supports forward cursors. Reverse cursors created with DECLARE SCROLL are not supported. cockroachdb/cockroach#77102
  • FETCH supports forward, relative, and absolute variants, but only for forward cursors. cockroachdb/cockroach#77102
  • BINARY CURSOR, which returns data in the Postgres binary format, is not supported. cockroachdb/cockroach#77099
  • WITH HOLD, which allows keeping a cursor open for longer than a transaction by writing its results into a buffer, is accepted as valid syntax within a single transaction but is not supported. It acts as a no-op and does not actually perform the function of WITH HOLD, which is to make the cursor live outside its parent transaction. Instead, if you are using WITH HOLD, you will be forced to close that cursor within the transaction it was created in. cockroachdb/cockroach#77101

    • This syntax is accepted (but does not have any effect):

      icon/buttons/copy

      BEGIN;
      DECLARE test_cur CURSOR WITH HOLD FOR SELECT * FROM foo ORDER BY bar;
      CLOSE test_cur;
      COMMIT;
      
    • This syntax is not accepted, and will result in an error:

      icon/buttons/copy

      BEGIN;
      DECLARE test_cur CURSOR WITH HOLD FOR SELECT * FROM foo ORDER BY bar;
      COMMIT; -- This will fail with an error because CLOSE test_cur was not called inside the transaction.
      
  • Scrollable cursor (also known as reverse FETCH) is not supported. cockroachdb/cockroach#77102

  • SELECT ... FOR UPDATE with a cursor is not supported. cockroachdb/cockroach#77103

  • Respect for SAVEPOINTs is not supported. Cursor definitions do not disappear properly if rolled back to a SAVEPOINT from before they were created. cockroachdb/cockroach#77104

SELECT FOR UPDATE locks are dropped on lease transfers and range splits/merges

By default under SERIALIZABLE isolation, locks acquired using SELECT ... FOR UPDATE and SELECT ... FOR SHARE are implemented as fast, in-memory unreplicated locks. If a lease transfer or range split/merge occurs on a range held by an unreplicated lock, the lock is dropped. The following behaviors can occur:

  • The desired ordering of concurrent accesses to one or more rows of a table expressed by your use of SELECT ... FOR UPDATE may not be preserved (that is, a transaction B against some table T that was supposed to wait behind another transaction A operating on T may not wait for transaction A).
  • The transaction that acquired the (now dropped) unreplicated lock may fail to commit, leading to transaction retry errors with code 40001 and the restart transaction error message.

When running under SERIALIZABLE isolation, SELECT ... FOR UPDATE and SELECT ... FOR SHARE locks should be thought of as best-effort, and should not be relied upon for correctness. Note that serialization is preserved despite this limitation. This limitation is fixed when the enable_durable_locking_for_serializable session setting is set to true.

Note:

This limitation does not apply to READ COMMITTED transactions.

Unsupported trigram syntax

The following PostgreSQL syntax and features are currently unsupported for trigrams:

  • word_similarity() built-in function.
  • strict_word_similarity() built-in function.
  • %> and <% comparisons and acceleration.
  • <<% and %>> comparisons and acceleration.
  • <->, <<->, <->>, <<<->, and <->>> comparisons.
  • Acceleration on regex string matching.
  • % comparisons, show_trgm, and trigram index creation on collated strings.

Tracking GitHub Issue

Statements containing multiple modification subqueries of the same table are disallowed

Statements containing multiple modification subqueries mutating the same row could cause corruption. These statements are disallowed by default, but you can enable multiple modification subqueries with one the following:

Note that if multiple mutations inside the same statement affect different tables with FOREIGN KEY relations and ON CASCADE clauses between them, the results will be different from what is expected in PostgreSQL.

Tracking GitHub Issue

transaction_rows_read_err and transaction_rows_written_err do not halt query execution

The transaction_rows_read_err and transaction_rows_written_err session settings limit the number of rows read or written by a single transaction. These session settings will fail the transaction with an error, but not until the current query finishes executing and the results have been returned to the client.

Tracking GitHub Issue

sql.guardrails.max_row_size_err misses indexed virtual computed columns

The sql.guardrails.max_row_size_err cluster setting misses large rows caused by indexed virtual computed columns. This is because the guardrail only checks the size of primary key rows, not secondary index rows.

Tracking GitHub Issue

CockroachDB does not allow inverted indexes with STORING

CockroachDB does not allow inverted indexes with a STORING column.

Tracking GitHub Issue

AS OF SYSTEM TIME does not support placeholders

CockroachDB does not support placeholders in AS OF SYSTEM TIME. The time value must be embedded in the SQL string.

Tracking GitHub issue

CockroachDB does not properly optimize some left and anti joins with GIN indexes

Left joins and anti joins involving JSONB, ARRAY, or spatial-typed columns with a multi-column or partitioned GIN index will not take advantage of the index if the prefix columns of the index are unconstrained, or if they are constrained to multiple, constant values.

To work around this limitation, make sure that the prefix columns of the index are either constrained to single constant values, or are part of an equality condition with an input column (e.g., col1 = col2, where col1 is a prefix column and col2 is an input column).

For example, suppose you have the following multi-region database and tables:

CREATE DATABASE multi_region_test_db PRIMARY REGION "europe-west1" REGIONS "us-west1", "us-east1" SURVIVE REGION FAILURE;
USE multi_region_test_db;

CREATE TABLE t1 (
  k INT PRIMARY KEY,
  geom GEOMETRY
);

CREATE TABLE t2 (
  k INT PRIMARY KEY,
  geom GEOMETRY,
  INVERTED INDEX geom_idx (geom)
) LOCALITY REGIONAL BY ROW;

And you insert some data into the tables:

INSERT INTO t1 SELECT generate_series(1, 1000), 'POINT(1.0 1.0)';
INSERT INTO t2 (crdb_region, k, geom) SELECT 'us-east1', generate_series(1, 1000), 'POINT(1.0 1.0)';
INSERT INTO t2 (crdb_region, k, geom) SELECT 'us-west1', generate_series(1001, 2000), 'POINT(2.0 2.0)';
INSERT INTO t2 (crdb_region, k, geom) SELECT 'europe-west1', generate_series(2001, 3000), 'POINT(3.0 3.0)';

If you attempt a left join between t1 and t2 on only the geometry columns, CockroachDB will not be able to plan an inverted join:

> EXPLAIN SELECT * FROM t1 LEFT JOIN t2 ON st_contains(t1.geom, t2.geom);
                info
------------------------------------
  distribution: full
  vectorized: true

  • cross join (right outer)
  │ pred: st_contains(geom, geom)
  │
  ├── • scan
  │     estimated row count: 3,000
  │     table: t2@primary
  │     spans: FULL SCAN
  │
  └── • scan
        estimated row count: 1,000
        table: t1@primary
        spans: FULL SCAN
(15 rows)

However, if you constrain the crdb_region column to a single value, CockroachDB can plan an inverted join:

> EXPLAIN SELECT * FROM t1 LEFT JOIN t2 ON st_contains(t1.geom, t2.geom) AND t2.crdb_region = 'us-east1';
                       info
--------------------------------------------------
  distribution: full
  vectorized: true

  • lookup join (left outer)
  │ table: t2@primary
  │ equality: (crdb_region, k) = (crdb_region,k)
  │ equality cols are key
  │ pred: st_contains(geom, geom)
  │
  └── • inverted join (left outer)
      │ table: t2@geom_idx
      │
      └── • render
          │
          └── • scan
                estimated row count: 1,000
                table: t1@primary
                spans: FULL SCAN
(18 rows)

If you do not know which region to use, you can combine queries with UNION ALL:

> EXPLAIN SELECT * FROM t1 LEFT JOIN t2 ON st_contains(t1.geom, t2.geom) AND t2.crdb_region = 'us-east1'
UNION ALL SELECT * FROM t1 LEFT JOIN t2 ON st_contains(t1.geom, t2.geom) AND t2.crdb_region = 'us-west1'
UNION ALL SELECT * FROM t1 LEFT JOIN t2 ON st_contains(t1.geom, t2.geom) AND t2.crdb_region = 'europe-west1';
                           info
----------------------------------------------------------
  distribution: full
  vectorized: true

  • union all
  │
  ├── • union all
  │   │
  │   ├── • lookup join (left outer)
  │   │   │ table: t2@primary
  │   │   │ equality: (crdb_region, k) = (crdb_region,k)
  │   │   │ equality cols are key
  │   │   │ pred: st_contains(geom, geom)
  │   │   │
  │   │   └── • inverted join (left outer)
  │   │       │ table: t2@geom_idx
  │   │       │
  │   │       └── • render
  │   │           │
  │   │           └── • scan
  │   │                 estimated row count: 1,000
  │   │                 table: t1@primary
  │   │                 spans: FULL SCAN
  │   │
  │   └── • lookup join (left outer)
  │       │ table: t2@primary
  │       │ equality: (crdb_region, k) = (crdb_region,k)
  │       │ equality cols are key
  │       │ pred: st_contains(geom, geom)
  │       │
  │       └── • inverted join (left outer)
  │           │ table: t2@geom_idx
  │           │
  │           └── • render
  │               │
  │               └── • scan
  │                     estimated row count: 1,000
  │                     table: t1@primary
  │                     spans: FULL SCAN
  │
  └── • lookup join (left outer)
      │ table: t2@primary
      │ equality: (crdb_region, k) = (crdb_region,k)
      │ equality cols are key
      │ pred: st_contains(geom, geom)
      │
      └── • inverted join (left outer)
          │ table: t2@geom_idx
          │
          └── • render
              │
              └── • scan
                    estimated row count: 1,000
                    table: t1@primary
                    spans: FULL SCAN
(54 rows)

Tracking GitHub Issue

Inverted join for tsvector and tsquery types is not supported

CockroachDB cannot index-accelerate queries with @@ predicates when both sides of the operator are variables.

Tracking GitHub issue

Using RESTORE with multi-region table localities

  • Restoring GLOBAL and REGIONAL BY TABLE tables into a non-multi-region database is not supported. Tracking GitHub Issue

  • REGIONAL BY TABLE and REGIONAL BY ROW tables can be restored only if the regions of the backed-up table match those of the target database. All of the following must be true for RESTORE to be successful:

    • The regions of the source database and the regions of the destination database have the same set of regions.
    • The regions were added to each of the databases in the same order.
    • The databases have the same primary region.

    The following example would be considered as having mismatched regions because the database regions were not added in the same order and the primary regions do not match.

    Running on the source database:

    ALTER DATABASE source_database SET PRIMARY REGION "us-east1";
    
    ALTER DATABASE source_database ADD region "us-west1";  
    

    Running on the destination database:

    ALTER DATABASE destination_database SET PRIMARY REGION "us-west1";
    
    ALTER DATABASE destination_database ADD region "us-east1";  
    

    In addition, the following scenario has mismatched regions between the databases since the regions were not added to the database in the same order.

    Running on the source database:

    ALTER DATABASE source_database SET PRIMARY REGION "us-east1";
    
    ALTER DATABASE source_database ADD region "us-west1";  
    

    Running on the destination database:

    ALTER DATABASE destination_database SET PRIMARY REGION "us-west1";
    
    ALTER DATABASE destination_database ADD region "us-east1";
    
    ALTER DATABASE destination_database SET PRIMARY REGION "us-east1";    
    

Tracking GitHub Issue

SET does not ROLLBACK in a transaction

SET does not properly apply ROLLBACK within a transaction. For example, in the following transaction, showing the TIME ZONE variable does not return 2 as expected after the rollback:

SET TIME ZONE +2;
BEGIN;
SET TIME ZONE +3;
ROLLBACK;
SHOW TIME ZONE;
timezone
------------
3

Tracking GitHub Issue

JSONB/JSON comparison operators are not implemented

You cannot use comparison operators (such as < or >) on JSONB elements. For example, the following query does not work and returns an error:

icon/buttons/copy

  SELECT '{"a": 1}'::JSONB -> 'a' < '{"b": 2}'::JSONB -> 'b';
  ERROR: unsupported comparison operator: <jsonb> < <jsonb>
  SQLSTATE: 22023

Tracking GitHub issue

Locality-optimized search works only for queries selecting a limited number of records

Expression indexes cannot reference computed columns

CockroachDB does not allow expression indexes to reference computed columns.

Tracking GitHub Issue

Materialized view limitations

  • The optimizer may not select the most optimal query plan when querying materialized views because CockroachDB does not collect statistics on materialized views.

    Tracking GitHub Issue.

  • CockroachDB cannot refresh materialized views inside explicit transactions. Trying to refresh a materialized view inside an explicit transaction will result in an error.

    1. Start cockroach demo with the sample bank data set:

      icon/buttons/copy
      cockroach demo bank
      
    2. Create the materialized view described in Usage.

    3. Start a new multi-statement transaction with BEGIN TRANSACTION:

      icon/buttons/copy

        BEGIN TRANSACTION;
      
    4. Inside the open transaction, attempt to refresh the view. This will result in an error.

      icon/buttons/copy

        REFRESH MATERIALIZED VIEW overdrawn_accounts;
      
        ERROR: cannot refresh view in an explicit transaction
        SQLSTATE: 25000
      

    Tracking GitHub Issue

CockroachDB cannot plan locality optimized searches that use partitioned unique indexes on virtual computed columns

Expressions as ON CONFLICT targets are not supported

CockroachDB does not support expressions as ON CONFLICT targets. This means that unique expression indexes cannot be selected as arbiters for INSERT .. ON CONFLICT statements. For example:

icon/buttons/copy
CREATE TABLE t (a INT, b INT, UNIQUE INDEX ((a + b)));
CREATE TABLE
icon/buttons/copy
INSERT INTO t VALUES (1, 2) ON CONFLICT ((a + b)) DO NOTHING;
invalid syntax: statement ignored: at or near "(": syntax error
SQLSTATE: 42601
DETAIL: source SQL:
INSERT INTO t VALUES (1, 2) ON CONFLICT ((a + b)) DO NOTHING
                                    ^
HINT: try \h INSERT
icon/buttons/copy
INSERT INTO t VALUES (1, 2) ON CONFLICT ((a + b)) DO UPDATE SET a = 10;
invalid syntax: statement ignored: at or near "(": syntax error
SQLSTATE: 42601
DETAIL: source SQL:
INSERT INTO t VALUES (1, 2) ON CONFLICT ((a + b)) DO UPDATE SET a = 10
                                    ^
HINT: try \h INSERT

Tracking GitHub Issue

Automatic statistics refresher may not refresh after upgrade

The automatic statistics refresher automatically checks whether it needs to refresh statistics for every table in the database upon startup of each node in the cluster. If statistics for a table have not been refreshed in a while, this will trigger collection of statistics for that table. If statistics have been refreshed recently, it will not force a refresh. As a result, the automatic statistics refresher does not necessarily perform a refresh of statistics after an upgrade. This could cause a problem, for example, if the upgrade moves from a version without histograms to a version with histograms. To refresh statistics manually, use CREATE STATISTICS.

Tracking GitHub Issue

Collection of statistics for virtual computed columns

CockroachDB does not collect statistics for virtual computed columns. This can prevent the optimizer from accurately calculating the cost of scanning an index on a virtual column, and, transitively, the cost of scanning an expression index.

Tracking GitHub issue

Differences in syntax and behavior between CockroachDB and PostgreSQL

CockroachDB supports the PostgreSQL wire protocol and the majority of its syntax. However, CockroachDB does not support some of the PostgreSQL features or behaves differently from PostgreSQL because not all features can be easily implemented in a distributed system.

For a list of known differences in syntax and behavior between CockroachDB and PostgreSQL, see Features that differ from PostgreSQL.

Multiple arbiter indexes for INSERT ON CONFLICT DO UPDATE

CockroachDB does not currently support multiple arbiter indexes for INSERT ON CONFLICT DO UPDATE, and will return an error if there are multiple unique or exclusion constraints matching the ON CONFLICT DO UPDATE specification.

Tracking GitHub Issue

Spatial support limitations

CockroachDB supports efficiently storing and querying spatial data, with the following limitations:

Limitations for composite types

Enterprise BACKUP does not capture database/table/column comments

The COMMENT ON statement associates comments to databases, tables, or columns. However, the internal table (system.comments) in which these comments are stored is not captured by a BACKUP of a table or database.

As a workaround, take a cluster backup instead, as the system.comments table is included in cluster backups.

Tracking GitHub Issue

DB Console may become inaccessible for secure clusters

Accessing the DB Console for a secure cluster now requires login information (i.e., username and password). This login information is stored in a system table that is replicated like other data in the cluster. If a majority of the nodes with the replicas of the system table data go down, users will be locked out of the DB Console.

Using LIKE...ESCAPE in WHERE and HAVING constraints

CockroachDB tries to optimize most comparisons operators in WHERE and HAVING clauses into constraints on SQL indexes by only accessing selected rows. This is done for LIKE clauses when a common prefix for all selected rows can be determined in the search pattern (e.g., ... LIKE 'Joe%'). However, this optimization is not yet available if the ESCAPE keyword is also used.

Tracking GitHub Issue

Tracking GitHub Issue

Current sequence value not checked when updating min/max value

Altering the minimum or maximum value of a series does not check the current value of a series. This means that it is possible to silently set the maximum to a value less than, or a minimum value greater than, the current value.

Tracking GitHub Issue

Using default_int_size session variable in batch of statements

When setting the default_int_size session variable in a batch of statements such as SET default_int_size='int4'; SELECT 1::IN, the default_int_size variable will not take affect until the next statement. This happens because statement parsing takes place asynchronously from statement execution.

As a workaround, set default_int_size via your database driver, or ensure that SET default_int_size is in its own statement.

Tracking GitHub Issue

COPY syntax not supported by CockroachDB

CockroachDB does not yet support the following COPY syntax:

Import with a high amount of disk contention

IMPORT can sometimes fail with a "context canceled" error, or can restart itself many times without ever finishing. If this is happening, it is likely due to a high amount of disk contention. This can be mitigated by setting the kv.bulk_io_write.max_rate cluster setting to a value below your max disk write speed. For example, to set it to 10MB/s, execute:

icon/buttons/copy
> SET CLUSTER SETTING kv.bulk_io_write.max_rate = '10MB';

Placeholders in PARTITION BY

When defining a table partition, either during table creation or table alteration, it is not possible to use placeholders in the PARTITION BY clause.

Tracking GitHub Issue

Dropping a single partition

CockroachDB does not currently support dropping a single partition from a table. In order to remove partitions, you can repartition the table.

Adding a column with sequence-based DEFAULT values

It is currently not possible to add a column to a table when the column uses a sequence as the DEFAULT value, for example:

icon/buttons/copy
> CREATE TABLE t (x INT);
icon/buttons/copy
> INSERT INTO t(x) VALUES (1), (2), (3);
icon/buttons/copy
> CREATE SEQUENCE s;
icon/buttons/copy
> ALTER TABLE t ADD COLUMN y INT DEFAULT nextval('s');
ERROR: nextval(): unimplemented: cannot evaluate scalar expressions containing sequence operations in this context
SQLSTATE: 0A000

Tracking GitHub Issue

Available capacity metric in the DB Console

If you are testing your deployment locally with multiple CockroachDB nodes running on a single machine (this is not recommended in production), you must explicitly set the store size per node in order to display the correct capacity. Otherwise, the machine's actual disk capacity will be counted as a separate store for each node, thus inflating the computed capacity.

Schema changes within transactions

Within a single transaction:

Note:

If a schema change within a transaction fails, manual intervention may be needed to determine which statement has failed. After determining which schema change(s) failed, you can then retry the schema change.

Schema change DDL statements inside a multi-statement transaction can fail while other statements succeed

Most schema change DDL statements that run inside a multi-statement transaction with non-DDL statements can fail at COMMIT time, even if other statements in the transaction succeed. This leaves such transactions in a "partially committed, partially aborted" state that may require manual intervention to determine whether the DDL statements succeeded.

Some DDL statements do not have this limitation. CREATE TABLE and CREATE INDEX statements have the same atomicity guarantees as other statements within a transaction.

If such a failure occurs, CockroachDB will emit a CockroachDB-specific error code, XXA00, and the following error message:

transaction committed but schema change aborted with error: <description of error>
HINT: Some of the non-DDL statements may have committed successfully, but some of the DDL statement(s) failed.
Manual inspection may be required to determine the actual state of the database.
Warning:

If you must execute schema change DDL statements inside a multi-statement transaction, we strongly recommend checking for this error code and handling it appropriately every time you execute such transactions.

This error will occur in various scenarios, including but not limited to:

  • Creating a unique index fails because values aren't unique.
  • The evaluation of a computed value fails.
  • Adding a constraint (or a column with a constraint) fails because the constraint is violated for the default/computed values in the column.

To see an example of this error, start by creating the following table.

icon/buttons/copy
CREATE TABLE T(x INT);
INSERT INTO T(x) VALUES (1), (2), (3);

Then, enter the following multi-statement transaction, which will trigger the error.

icon/buttons/copy
BEGIN;
ALTER TABLE t ADD CONSTRAINT unique_x UNIQUE(x);
INSERT INTO T(x) VALUES (3);
COMMIT;
pq: transaction committed but schema change aborted with error: (23505): duplicate key value (x)=(3) violates unique constraint "unique_x"
HINT: Some of the non-DDL statements may have committed successfully, but some of the DDL statement(s) failed.
Manual inspection may be required to determine the actual state of the database.

In this example, the INSERT statement committed, but the ALTER TABLE statement adding a UNIQUE constraint failed. We can verify this by looking at the data in table t and seeing that the additional non-unique value 3 was successfully inserted.

icon/buttons/copy
SELECT * FROM t;
  x
+---+
  1
  2
  3
  3
(4 rows)

Schema changes between executions of prepared statements

When the schema of a table targeted by a prepared statement changes after the prepared statement is created, future executions of the prepared statement could result in an error. For example, adding a column to a table referenced in a prepared statement with a SELECT * clause will result in an error:

icon/buttons/copy
CREATE TABLE users (id INT PRIMARY KEY);
icon/buttons/copy
PREPARE prep1 AS SELECT * FROM users;
icon/buttons/copy
ALTER TABLE users ADD COLUMN name STRING;
icon/buttons/copy
INSERT INTO users VALUES (1, 'Max Roach');
icon/buttons/copy
EXECUTE prep1;
ERROR: cached plan must not change result type
SQLSTATE: 0A000

It's therefore recommended to explicitly list result columns instead of using SELECT * in prepared statements, when possible.

Declarative schema changer does not track rows in system.privileges

The declarative schema changer does not track rows in the system.privileges table, which prevents the declarative schema changer from successfully running the DROP OWNED BY statement.

Tracking GitHub issue

Size limits on statement input from SQL clients

CockroachDB imposes a hard limit of 16MiB on the data input for a single statement passed to CockroachDB from a client (including the SQL shell). We do not recommend attempting to execute statements from clients with large input.

Using \| to perform a large input in the SQL shell

In the built-in SQL shell, using the \| operator to perform a large number of inputs from a file can cause the server to close the connection. This is because \| sends the entire file as a single query to the server, which can exceed the upper bound on the size of a packet the server can accept from any client (16MB).

As a workaround, execute the file from the command line with cat data.sql | cockroach sql instead of from within the interactive shell.

New values generated by DEFAULT expressions during ALTER TABLE ADD COLUMN

When executing an ALTER TABLE ADD COLUMN statement with a DEFAULT expression, new values generated:

  • use the default search path regardless of the search path configured in the current session via SET SEARCH_PATH.
  • use the UTC time zone regardless of the time zone configured in the current session via SET TIME ZONE.
  • have no default database regardless of the default database configured in the current session via SET DATABASE, so you must specify the database of any tables they reference.
  • use the transaction timestamp for the statement_timestamp() function regardless of the time at which the ALTER statement was issued.

Load-based lease rebalancing in uneven latency deployments

When nodes are started with the --locality flag, CockroachDB attempts to place the replica lease holder (the replica that client requests are forwarded to) on the node closest to the source of the request. This means as client requests move geographically, so too does the replica lease holder.

However, you might see increased latency caused by a consistently high rate of lease transfers between datacenters in the following case:

  • Your cluster runs in datacenters which are very different distances away from each other.
  • Each node was started with a single tier of --locality, e.g., --locality=datacenter=a.
  • Most client requests get sent to a single datacenter because that's where all your application traffic is.

To detect if this is happening, open the DB Console, select the Queues dashboard, hover over the Replication Queue graph, and check the Leases Transferred / second data point. If the value is consistently larger than 0, you should consider stopping and restarting each node with additional tiers of locality to improve request latency.

For example, let's say that latency is 10ms from nodes in datacenter A to nodes in datacenter B but is 100ms from nodes in datacenter A to nodes in datacenter C. To ensure A's and B's relative proximity is factored into lease holder rebalancing, you could restart the nodes in datacenter A and B with a common region, --locality=region=foo,datacenter=a and --locality=region=foo,datacenter=b, while restarting nodes in datacenter C with a different region, --locality=region=bar,datacenter=c.

Overload resolution for collated strings

Many string operations are not properly overloaded for collated strings, for example:

icon/buttons/copy
> SELECT 'string1' || 'string2';
     ?column?
------------------
  string1string2
(1 row)
icon/buttons/copy
> SELECT ('string1' collate en) || ('string2' collate en);
pq: unsupported binary operator: <collatedstring{en}> || <collatedstring{en}>

Tracking GitHub Issue

Max size of a single column family

When creating or updating a row, if the combined size of all values in a single column family exceeds the max range size for the table, the operation may fail, or cluster performance may suffer.

As a workaround, you can either manually split a table's columns into multiple column families, or you can create a table-specific zone configuration with an increased max range size.

Simultaneous client connections and running queries on a single node

When a node has both a high number of client connections and running queries, the node may crash due to memory exhaustion. This is due to CockroachDB not accurately limiting the number of clients and queries based on the amount of available RAM on the node.

To prevent memory exhaustion, monitor each node's memory usage and ensure there is some margin between maximum CockroachDB memory usage and available system RAM. For more details about memory usage in CockroachDB, see this blog post.

To control the maximum number of non-superuser (root user or other admin role) connections a gateway node can have open at one time, use the server.max_connections_per_gateway cluster setting. If a new non-superuser connection would exceed this limit, the error message "sorry, too many clients already" is returned, along with error code 53300. This may be useful in addition to your memory monitoring.

Privileges for DELETE and UPDATE

Every DELETE or UPDATE statement constructs a SELECT statement, even when no WHERE clause is involved. As a result, the user executing DELETE or UPDATE requires both the DELETE and SELECT or UPDATE and SELECT privileges on the table.

ROLLBACK TO SAVEPOINT in high-priority transactions containing DDL

Transactions with priority HIGH that contain DDL and ROLLBACK TO SAVEPOINT are not supported, as they could result in a deadlock. For example:

> BEGIN PRIORITY HIGH; SAVEPOINT s; CREATE TABLE t(x INT); ROLLBACK TO SAVEPOINT s;
ERROR: unimplemented: cannot use ROLLBACK TO SAVEPOINT in a HIGH PRIORITY transaction containing DDL
SQLSTATE: 0A000
HINT: You have attempted to use a feature that is not yet implemented.
See: https://github.com/cockroachdb/cockroach/issues/46414

Tracking GitHub Issue

CockroachDB does not test for all connection failure scenarios

CockroachDB servers rely on the network to report when a TCP connection fails. In most scenarios when a connection fails, the network immediately reports a connection failure, resulting in a Connection refused error.

However, if there is no host at the target IP address, or if a firewall rule blocks traffic to the target address and port, a TCP handshake can linger while the client network stack waits for a TCP packet in response to network requests. To work around this kind of scenario, we recommend the following:

  • When migrating a node to a new machine, keep the server listening at the previous IP address until the cluster has completed the migration.
  • Configure any active network firewalls to allow node-to-node traffic.
  • Verify that orchestration tools (e.g., Kubernetes) are configured to use the correct network connection information.

Tracking GitHub Issue

Some column-dropping schema changes do not roll back properly

Some schema changes that drop columns cannot be rolled back properly.

In some cases, the rollback will succeed, but the column data might be partially or totally missing, or stale due to the asynchronous nature of the schema change.

Tracking GitHub Issue

In other cases, the rollback will fail in such a way that will never be cleaned up properly, leaving the table descriptor in a state where no other schema changes can be run successfully.

Tracking GitHub Issue

To reduce the chance that a column drop will roll back incorrectly:

  • Perform column drops in transactions separate from other schema changes. This ensures that other schema change failures will not cause the column drop to be rolled back.

  • Drop all constraints (including unique indexes) on the column in a separate transaction, before dropping the column.

  • Drop any default values or computed expressions on a column before attempting to drop the column. This prevents conflicts between constraints and default/computed values during a column drop rollback.

If you think a rollback of a column-dropping schema change has occurred, check the jobs table. Schema changes with an error prefaced by cannot be reverted, manual cleanup may be required might require manual intervention.

No guaranteed state switch from DECOMMISSIONING to DECOMMISSIONED if node decommission is interrupted

There is no guaranteed state switch from DECOMMISSIONING to DECOMMISSIONED if node decommission is interrupted in one of the following ways:

  • The cockroach node decommission --wait-all command was run and then interrupted
  • The cockroach node decommission --wait=none command was run

This is because the state flip is effected by the CLI program at the end. Only the CLI (or its underlying API call) is able to finalize the "decommissioned" state. If the command is interrupted, or --wait=none is used, the state will only flip to "decommissioned" when the CLI program is run again after decommissioning has done all its work.

Tracking GitHub issue

Remove a UNIQUE index created as part of CREATE TABLE

UNIQUE indexes created as part of a CREATE TABLE statement cannot be removed without using CASCADE. Unique indexes created with CREATE INDEX do not have this limitation.

Row-Level TTL limitations

  • Any queries you run against tables with Row-Level TTL enabled do not filter out expired rows from the result set (this includes UPDATEs and DELETEs). This feature may be added in a future release. For now, follow the instructions in Filter out expired rows from a selection query.
  • Enabling Row-Level TTL on a table with multiple secondary indexes can have negative performance impacts on a cluster, including increased latency and contention. This is particularly true for large tables with terabytes of data and billions of rows that are split up into multiple ranges across separate nodes.
    • Increased latency may occur because secondary indexes aren't necessarily stored on the same underlying ranges as a table's primary indexes. Further, the secondary indexes' ranges may have leaseholders located on different nodes than the primary index.
    • Increased contention may occur because intents must be written as part of performing the deletions.
    • Finally, secondary indexes can also have a negative impact on the overall performance of TTL jobs. According to internal testing, the TTL job processing rate is worse on tables with secondary indexes. If you encounter this situation, decreasing the ttl_delete_batch_size storage parameter may help by decreasing the number of ranges that need to be accessed by the job.

Change data capture limitations

Change data capture (CDC) provides efficient, distributed, row-level changefeeds into Apache Kafka for downstream processing such as reporting, caching, or full-text indexing. It has the following known limitations:


Yes No
On this page

Yes No