---
title: "Engines and versions"
description: "The engines and versions a database can run, its instance type and storage, and the engine parameters you can set."
---

> Documentation Index
> Fetch the complete documentation index at: https://docs.ezghcloud.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Engines and versions

| Engine | `engine` | Versions | Port |
| --- | --- | --- | --- |
| PostgreSQL | `postgresql` | 17, 18 | 5432 |
| Percona Server for MySQL | `percona` | 8.4 | 3306 |
| MariaDB Community Server | `mariadb` | 11.8, 12.3 | 3306 |

`ezgh databases engine-versions` ([ListEngineVersions](/databases-api/ListEngineVersions/))
lists them, with each version's parameters and extensions.

A database keeps its engine. Minor versions are applied for you in its
[maintenance window](/databases#maintenance). A new major version is a
[blue/green deployment](/databases/blue-green).

On Percona Server and MariaDB, tables use InnoDB: other storage engines are refused.

## Instance type

A database runs on a `gp-1` or `gs-1` [instance type](/compute/instance-types), for example
`gp-1.large`. Changing it applies at once, with one failover: open connections end, and clients
reconnect to the same endpoint.

## Storage

Storage is a `gp1` [volume type](/compute/volumes), with its size in GiB and, optionally, its IOPS
and throughput (the volume type's baseline by default).

Storage grows while the database runs, and never shrinks:

```sh
ezgh databases update dbi_4k2m9x0a7qpe --storage-size-gib 200 --wait
```

When storage fills up, the database's status is `storage_full` and writes fail. Growing it works in
that status too.

## Parameters

Parameters are set per database, as `name=value` strings, from the lists below. Anything else is
refused. A parameter you don't set keeps its default, which may be sized to the instance type and
storage.

```sh
ezgh databases update dbi_4k2m9x0a7qpe --parameter work_mem=65536 --parameter statement_timeout=30000
ezgh databases update dbi_4k2m9x0a7qpe --unset-parameter work_mem
```

In the API, [UpdateDatabase](/databases-api/UpdateDatabase/) takes the whole set: a parameter
left out goes back to its default.

A parameter marked **At once** applies with no interruption. One marked **Restart** restarts the
engine: open connections end, and clients reconnect to the same endpoint. Raise and lower
PostgreSQL's `max_connections`, `max_worker_processes`, `max_prepared_transactions` and
`max_locks_per_transaction` in separate changes.

Values are whole numbers, decimals, `on` or `off`, or one of the listed words, with no unit. A number
is in the unit the engine's documentation gives for a value without one: for example kB for
PostgreSQL's `work_mem`, milliseconds for its `statement_timeout`, and bytes for Percona Server's and
MariaDB's buffer sizes.

### Thread pool

On Percona Server, the thread pool is off: `thread_handling` is `one-thread-per-connection`. Set it
to `pool-of-threads` to turn the thread pool on. On MariaDB, `thread_handling` is `pool-of-threads`
unless you set it.

### PostgreSQL

| Parameter | Values | Applies |
| --- | --- | --- |
| `max_connections` | 25 to 5000 | Restart |
| `max_worker_processes` | 8 to 64 | Restart |
| `max_prepared_transactions` | 0 to 1000 | Restart |
| `max_locks_per_transaction` | 64 to 4096 | Restart |
| `work_mem` | 64 to 2,097,152 | At once |
| `maintenance_work_mem` | 1024 to 2,097,152 | At once |
| `autovacuum_work_mem` | -1 to 2,097,152 | At once |
| `temp_buffers` | 100 to 1,048,576 (8 kB blocks) | At once |
| `max_wal_size` | 80 to 65,536 | At once |
| `min_wal_size` | 32 to 8192 | At once |
| `checkpoint_completion_target` | 0.1 to 1 | At once |
| `random_page_cost` | 0.1 to 100 | At once |
| `effective_io_concurrency` | 0 to 1000 | At once |
| `default_statistics_target` | 1 to 10,000 | At once |
| `statement_timeout` | 0 to 86,400,000 | At once |
| `lock_timeout` | 0 to 86,400,000 | At once |
| `idle_in_transaction_session_timeout` | 0 to 86,400,000 | At once |
| `idle_session_timeout` | 0 to 86,400,000 | At once |
| `deadlock_timeout` | 1 to 60,000 | At once |
| `log_min_duration_statement` | -1 to 86,400,000 | At once |
| `log_autovacuum_min_duration` | -1 to 86,400,000 | At once |
| `log_temp_files` | -1 to 1,048,576 | At once |
| `log_lock_waits` | `on`, `off` | At once |
| `log_connections` | `on`, `off` | At once |
| `log_disconnections` | `on`, `off` | At once |
| `autovacuum` | `on`, `off` | At once |
| `autovacuum_max_workers` | 1 to 16 | Restart |
| `autovacuum_naptime` | 1 to 86,400 | At once |
| `autovacuum_vacuum_threshold` | 0 to 1,000,000 | At once |
| `autovacuum_vacuum_scale_factor` | 0 to 1 | At once |
| `autovacuum_analyze_threshold` | 0 to 1,000,000 | At once |
| `autovacuum_analyze_scale_factor` | 0 to 1 | At once |
| `autovacuum_vacuum_cost_limit` | -1 to 10,000 | At once |
| `autovacuum_vacuum_cost_delay` | -1 to 100 | At once |

`log_min_duration_statement` takes effect when the database [exports](/databases/monitoring#logs) its
`slow` log.

### Percona Server and MariaDB

| Parameter | Values | Applies |
| --- | --- | --- |
| `max_connections` | 25 to 10,000 | At once |
| `max_user_connections` | 0 to 10,000 | At once |
| `wait_timeout` | 1 to 86,400 | At once |
| `interactive_timeout` | 1 to 86,400 | At once |
| `lock_wait_timeout` | 1 to 86,400 | At once |
| `innodb_lock_wait_timeout` | 1 to 3600 | At once |
| `net_read_timeout` | 1 to 3600 | At once |
| `net_write_timeout` | 1 to 3600 | At once |
| `sort_buffer_size` | 32,768 to 1,073,741,824 | At once |
| `join_buffer_size` | 8192 to 1,073,741,824 | At once |
| `read_buffer_size` | 8192 to 1,073,741,824 | At once |
| `read_rnd_buffer_size` | 8192 to 1,073,741,824 | At once |
| `tmp_table_size` | 1024 to 17,179,869,184 | At once |
| `max_heap_table_size` | 16,384 to 17,179,869,184 | At once |
| `max_allowed_packet` | 1024 to 1,073,741,824 | At once |
| `table_open_cache` | 400 to 100,000 | At once |
| `thread_cache_size` | 0 to 16,384 | At once |
| `long_query_time` | 0 to 3600 | At once |
| `log_queries_not_using_indexes` | `on`, `off` | At once |
| `innodb_io_capacity` | 100 to 100,000 | At once |
| `innodb_io_capacity_max` | 100 to 100,000 | At once |
| `innodb_lru_scan_depth` | 100 to 10,000 | At once |
| `innodb_stats_persistent_sample_pages` | 1 to 10,000 | At once |
| `innodb_log_buffer_size` | 1,048,576 to 4,294,967,296 | Restart |
| `innodb_purge_threads` | 1 to 32 | Restart |
| `innodb_read_io_threads` | 1 to 64 | Restart |
| `innodb_write_io_threads` | 1 to 64 | Restart |
| `thread_handling` | `one-thread-per-connection`, `pool-of-threads` | Restart |
| `max_execution_time` (Percona Server) | 0 to 86,400,000 | At once |
| `max_statement_time` (MariaDB) | 0 to 86,400 | At once |

Source: https://docs.ezghcloud.com/databases/engines-and-versions/index.mdx
