memory_global_total different to mysql process -> kiling pod
Hi community, We are selfhosting a mariadb database in a kubernetes cluster. Since very recently the database runs out of memory regularly and restarts the kuberentes pod.
The problem is that sys.memory_global_total shows total_allocated around 5.5 GB but the actual memory usage of the mysqld process that the bash top command shows goes up to 20GB.
We are wondering why it even gets that high.
According to our configuration it should use far less than 20GB: binlog_cache_size 32.00 KiB innodb_buffer_pool_size 4.00 GiB innodb_log_buffer_size 16.00 MiB join_buffer_size 256.00 KiB key_buffer_size 10.00 MiB max_connections 500 query_cache_size 0 read_buffer_size 128.00 KiB read_rnd_buffer_size 256.00 KiB sort_buffer_size 2.00 MiB thread_stack 292.00 KiB tmp_table_size 2.00 MiB
Total Max Memory 6.4382 GB
Thanks for any input in advance.