🐞 Issue Report: MySQL Docker Container in Kubernetes Restarts Every 2–3 Days Despite 11 GB RAM Assigned to Pod

[jayrajmeh0]

Description

We're running a MySQL Docker container in a Kubernetes cluster. The container restarts every 2–3 days, typically when memory usage increases. We've assigned 11 GB of RAM directly to the Pod, yet the MySQL container still crashes and restarts periodically.

Interestingly, the same application and MySQL configuration runs fine on a Windows Server with only 8 GB RAM, without restarts.

We’ve already tuned my.cnf parameters to better manage memory usage, but that hasn't resolved the issue. We suspect the MySQL container may be hitting memory limits or being killed (possibly due to OOM conditions).

---

📦 Environment Details

• MySQL Version: e.g., 8.0.35
• Docker Image Used: e.g., mysql:8.0.35
• Kubernetes Version: e.g., v1.28
• Pod RAM Limit: 11 GB
• Windows Server (reference setup): 8 GB RAM — No restart issues
• Storage Type: _e.g., PVC / Rook-ceph
• Platform: Kubernetes

---

⚙️ MySQL Configuration (my.cnf)

[mysqld]
log_bin_trust_function_creators=1
sql-mode=ONLY_FULL_GROUP_BY,STRICT_TRANS_TABLES,NO_ZERO_IN_DATE,NO_ZERO_DATE,ERROR_FOR_DIVISION_BY_ZERO,NO_ENGINE_SUBSTITUTION
cte_max_recursion_depth=10000
group_concat_max_len=1000000
innodb_print_all_deadlocks=1
transaction_isolation=READ-COMMITTED
lower_case_table_names=1
default_authentication_plugin=mysql_native_password
local_infile=1
default-storage-engine=INNODB
slow-query-log=1
slow_query_log_file=/var/lib/mysql/ADMIN-slow.log
long_query_time=10
log-error=/var/lib/mysql/ADMIN.err
innodb_buffer_pool_size=512M
tmp_table_size=64M
thread_cache_size=10
table_open_cache=500
max_connections=100
key_buffer_size=64M
read_buffer_size=4M
read_rnd_buffer_size=4M
innodb_log_buffer_size=8M
innodb_log_file_size=48M
innodb_thread_concurrency=25
innodb_buffer_pool_instances=4
back_log=80
join_buffer_size=8M
max_allowed_packet=64M
open_files_limit=8161
sort_buffer_size=4M
connect_timeout=1000
max_heap_table_size=40M
innodb_log_files_in_group=8
max_length_for_sort_data=16K
innodb_print_all_deadlocks=1
innodb_ft_result_cache_limit=128M
innodb_ft_total_cache_size=64M
thread_stack=512K
wait_timeout=60
interactive_timeout=60
performance_schema=ON

[LaurentGoderre]

Are you sure it's memory related and not file system related (for example, running out of space for logs)?

[jayrajmeh0]

@LaurentGoderre Thanks for your quick response.

My MySQL container starts by using around 25% of the allocated memory, but after running for about 21 hours, its usage gradually increases to approximately 56%. Currently, the pod is allocated 11 GB of RAM, and it tends to restart every 3 days due to memory limits. If I increase the memory to 16 GB, it may extend the uptime to around 5 days before a restart. However, simply increasing memory is not a long-term or ideal solution.

What’s more interesting is that the same project is also running on an IIS server with MySQL on Windows, which only has 8 GB of RAM. That instance has been running for nearly 2 months without any restarts or issues.

I hope this clarifies my concern — the problem seems to arise specifically when using the MySQL Docker image running as a container in Kubernetes The pod has a memory limit of 11 GB, while the node has a total of 16 GB RAM.

[jayrajmeh0]

@LaurentGoderre Thanks for your quick response.

My MySQL container starts by using around 25% of the allocated memory, but after running for about 21 hours, its usage gradually increases to approximately 56%. Currently, the pod is allocated 11 GB of RAM, and it tends to restart every 3 days due to memory limits. If I increase the memory to 16 GB, it may extend the uptime to around 5 days before a restart. However, simply increasing memory is not a long-term or ideal solution.

What’s more interesting is that the same project is also running on an IIS server wi

@LaurentGoderre @thaJeztah @woutersamaey @tianon Does anyone know the solution to this issue? If yes, please explain.

[thaJeztah]

It could still be related to ulimits set for the container in which MySQL runs; see the links I posted in my other comment; #1112 (comment)

MySQL, and various other tools (as outlined in some of those tickets) may adjust itself to such limits. If the container has (very) high limits, it may adjust itself to that, and use more resources. The solution to that could be to adjust those limits for the container to limits suitable for your use.

[thaJeztah]

Some pointers are also given here; https://docs.docker.com/engine/release-notes/25.0/#new

[thaJeztah]

Might also want to check what sysctls are set for the pod/container; see

• incredible slowness for yum operation into a centos* container moby/moby#45838 (comment)
• 1.5.10 causes memory leak in mysql container; regression from 1.4.13 containerd/containerd#6707