TECH / Servers & Websites 2 min read

How to Fix “MySQL server has gone away” (Error 2006) – Causes & Solutions

Antonio NEMCD Author
Published Mar 19, 2012 · 2 min

The MySQL server has gone away error (code 2006) is one of the most common headaches when importing large SQL dumps, running WordPress migrations, or executing long-running PHP/Python scripts. It means your client sent a query after MySQL or MariaDB had already closed the connection.

Fixing MySQL server has gone away error 2006

Quick answer: 90% of the time it comes down to two server settings – max_allowed_packet being too small (large query or SQL dump import) or wait_timeout expiring on an idle connection.

Step 1. Check Current Server Variables

Run these queries in the MySQL/MariaDB CLI or phpMyAdmin’s SQL tab:

SHOW VARIABLES LIKE 'max_allowed_packet';
SHOW VARIABLES LIKE 'wait_timeout';
SHOW VARIABLES LIKE 'interactive_timeout';

Default max_allowed_packet is often just 16 MB (or 1–4 MB on older hosts). If a query or bulk insert exceeds that limit, MySQL treats the packet as malformed and drops the connection immediately.

Step 2. Increase Limits in my.cnf / mysqld.cnf

Open your database config file (on Ubuntu/Debian usually /etc/mysql/mysql.conf.d/mysqld.cnf or /etc/mysql/mariadb.conf.d/50-server.cnf; on CentOS/Rocky /etc/my.cnf) and raise these values under [mysqld]:

[mysqld]
max_allowed_packet = 64M
wait_timeout = 600
interactive_timeout = 600

Restart the database service:

sudo systemctl restart mysql
# or for MariaDB:
sudo systemctl restart mariadb

Step 3. Importing Large SQL Dumps via CLI

When restoring a big backup from the command line, pass an increased packet limit to the mysql client as well:

mysql --max_allowed_packet=256M -u root -p dbname < backup.sql

2 Other Frequent Causes of Error 2006

  • Out-of-Memory (OOM Killer). On low-RAM VPS plans, Linux can kill mysqld under load spikes, leaving clients with MySQL server has gone away. Check dmesg | grep -i oom, add swap space, or move to a roomier VPS (such as Fornex with 7% off or Vultr – affiliate links).
  • Application logic. A worker opens a DB connection, spends 30 minutes downloading files or scraping pages without querying SQL, and then reuses a timed-out handle (or queries a handle it already closed). Add a ping/reconnect check in long-running workers.

Related database guide: MariaDB vs MySQL – which to choose for your server.

Start by checking SHOW VARIABLES and dmesg for OOM events – bumping max_allowed_packet and wait_timeout fixes the vast majority of cases in two minutes.

Related posts:

Leave a Comment