ScyllaDB Docs ScyllaDB Operator Troubleshoot Collect debugging information Query system tables for debugging

Query system tables for debugging¶

Use ScyllaDB system tables to gather diagnostic information that is not captured by must-gather.

System tables contain ScyllaDB-internal state — repair history, SSTable activity, large partitions, runtime information, and more.

When to use system tables¶

Use case	Tool
Kubernetes-level diagnostics (pod status, events, logs)	Collect debugging information
ScyllaDB-internal state (repair history, SSTable stats, large partitions)	System tables (this page)

Include system table output alongside debugging information archives when filing support tickets for ScyllaDB-level issues (performance, data consistency, repair).

How to query system tables¶

Connect to a ScyllaDB pod and use cqlsh:

kubectl -n scylla exec -it <pod-name> -c scylla -- cqlsh

Or run a single query:

kubectl -n scylla exec -it <pod-name> -c scylla -- \
  cqlsh -e "<CQL query>"

Useful system tables¶

Cluster and node information¶

-- Node-local information (listen address, host ID, cluster name, ScyllaDB version)
SELECT * FROM system.local;

-- All known peers in the cluster
SELECT peer, data_center, rack, host_id, schema_version, tokens FROM system.peers;

Example output (system.local):

 key   | bootstrapped | cluster_name | cql_version | data_center
-------+--------------+--------------+-------------+-------------
 local | COMPLETED    | my-cluster   | 3.4.5       | us-east-1

Example output (system.peers):

 peer      | data_center | host_id                              | rack       | release_version | tokens
-----------+-------------+--------------------------------------+------------+-----------------+--------
 10.0.1.5  | us-east-1   | a1b2c3d4-e5f6-7890-abcd-ef1234567890 | us-east-1a | 2025.1.0        | {-9223372036854775808, ...}
 10.0.1.6  | us-east-1   | e5f6a7b8-c9d0-1234-5678-9abcdef01234 | us-east-1b | 2025.1.0        | {-1234567890123456789, ...}

Large partitions¶

-- Partitions that exceed the large partition threshold
SELECT * FROM system.large_partitions LIMIT 20;

-- Large rows
SELECT * FROM system.large_rows LIMIT 20;

-- Large cells
SELECT * FROM system.large_cells LIMIT 20;

Example output (system.large_partitions):

 keyspace_name | table_name | sstable_name    | partition_size | partition_key | rows
---------------+------------+-----------------+----------------+---------------+------
 mykeyspace    | mytable    | md-123-big-Data | 1048576        | 'mykey'       | 1000
 mykeyspace    | mytable    | md-124-big-Data | 524288         | 'otherkey'    | 500

Example output (system.large_rows):

 keyspace_name | table_name | sstable_name    | row_size | partition_key | clustering_key
---------------+------------+-----------------+----------+---------------+---------------
 mykeyspace    | mytable    | md-124-big-Data | 524288   | 'mykey'       | 'myrow'

Example output (system.large_cells):

 keyspace_name | table_name | sstable_name    | cell_size | partition_key | clustering_key | column_name
---------------+------------+-----------------+-----------+---------------+----------------+------------
 mykeyspace    | mytable    | md-125-big-Data | 262144    | 'mykey'       | 'myrow'        | 'mycolumn'

Large partitions are a common cause of performance issues and timeouts. See the ScyllaDB documentation on large partitions.

Repair history¶

-- Repair history (distributed across nodes)
SELECT * FROM system_distributed.repair_history LIMIT 20;

Example output (system_distributed.repair_history):

 id   | completed_at                    | keyspace_name | options | range_end | range_start | status  | table_names
------+---------------------------------+---------------+---------+-----------+-------------+---------+------------
 uuid | 2026-04-01 10:30:00.000000+0000 | mykeyspace    | {}      | ...       | ...         | SUCCESS | {'mytable'}
 uuid | 2026-04-01 09:15:00.000000+0000 | mykeyspace    | {}      | ...       | ...         | SUCCESS | {'mytable'}

SSTable activity¶

-- SSTable activity statistics
SELECT * FROM system.sstable_activity LIMIT 20;

Example output (system.sstable_activity):

 keyspace_name | columnfamily_name | generation | rate_120m | rate_15m
---------------+-------------------+------------+-----------+---------
 mykeyspace    | mytable           | 123        |       0.5 |      1.2
 mykeyspace    | mytable           | 124        |       0.2 |      0.8

Runtime information¶

-- Runtime configuration and memory usage
SELECT * FROM system.runtime_info;

Example output (system.runtime_info):

 group | item            | value
-------+-----------------+---------
 cache | entries         |     193
 cache | hit_rate_recent |      82
 cache | hit_rate_total  |    0.99
 cache | hits            |  122282
 cache | memory_free     | 5950484

Compaction history¶

-- Recent compaction events
SELECT * FROM system.compaction_history LIMIT 20;

Example output (system.compaction_history):

 id   | bytes_in | bytes_out | columnfamily_name | compaction_properties | ended_at                        | keyspace_name | rows_merged
------+----------+-----------+-------------------+-----------------------+---------------------------------+---------------+------------
 uuid | 10485760 |   8388608 | mytable           | {}                    | 2026-04-01 09:00:00.000000+0000 | mykeyspace    | {1: 5000}
 uuid |  5242880 |   4194304 | mytable           | {}                    | 2026-04-01 08:30:00.000000+0000 | mykeyspace    | {1: 2500}

Include system table output in support tickets¶

When filing a support ticket, include:

Output from system.local and system.peers — provides cluster topology context.
Output from system.large_partitions — if the issue is performance-related.
Output from system_distributed.repair_history — if the issue involves data consistency.
Any other tables relevant to the specific issue.

Format the output for readability:

kubectl -n scylla exec -it <pod-name> -c scylla -- \
  cqlsh -e "EXPAND ON; SELECT * FROM system.local;"

The EXPAND ON command formats each row vertically, making wide tables easier to read.

Was this page helpful?