The memory control group (memcg) backend asynchronous reclaim feature is provided in Alibaba Cloud Linux 2 kernel version 4.19.81-17.al7 or later. This topic describes the interface files that implement the memcg backend asynchronous reclaim feature.

Background information

In community versions of the Linux kernel, the system allocates memory and triggers memcg-level direct memory reclaim when the memory reaches a limit set by a memcg. Direct memory reclaim is synchronous reclaim that occurs in the context of memory allocation and affects performance of the current process.

To resolve this problem, Alibaba Cloud Linux 2 provides the backend asynchronous reclaim feature for memcgs. This feature differs from the global kswapd kernel thread in that it uses the workqueue mechanism instead of a corresponding memcg kswapd kernel thread. Additionally, four memcg interface files are added in each of the cgroup v1 and cgroup v2 interfaces.

Precautions
  • Memory allocation of the current memcg may recursively trigger the backend asynchronous reclaim of the parent group.
  • When triggered, backend asynchronous reclaim starts from the memcg on which the feature is triggered and is performed in sequence down the hierarchy.
  • When memory.high is configured and the value of memory.high is smaller than that of memory.limit_in_bytes, the values of memory.wmark_high and memory.wmark_low are calculated based on memory.high instead of memory.limit_in_bytes.

Interface files

Interface file Description
memory.wmark_ratio Specifies whether to enable the memcg backend asynchronous reclaim feature and sets the memcg memory watermark where asynchronous reclaim starts. Unit: percent of the memcg memory limit. Valid values: 0 to 100.
  • The default value is 0, which indicates that the memcg backend asynchronous reclaim feature is disabled.
  • When the value is not 0, the memcg backend asynchronous reclaim feature is enabled and the corresponding watermark is set.
memory.wmark_high A read-only interface file.
  • When the memcg memory usage exceeds the value of this interface file, backend asynchronous reclaim is started.
  • The value of this interface file is calculated based on the following formula: (memory.limit_in_bytes × memory.wmark_ratio/100).
  • When the memcg backend asynchronous reclaim feature is disabled, the default value of memory.wmark_high is the maximum value and the backend asynchronous reclaim feature is never triggered.
  • This interface file is not stored in the memcg root directory.
memory.wmark_low A read-only interface file.
  • When the memcg memory usage is less than the value of this interface file, backend asynchronous reclaim is ended.
  • The value of this interface file is calculated based on the following formula: memory.wmark_high - memory.limit_in_bytes × memory.wmark_scale_factor/10000.
  • This interface file is not stored in the memcg root directory.
memory.wmark_scale_factor Controls the gap between memory.wmark_high and memory.wmark_low. Unit: 0.01% of the memcg memory limit. Valid values: 1 to 1000.
  • This interface file inherits the value of its parent group when the interface file is created. The value of this interface file is 50, which indicates 0.50% of the memcg memory limit. This is the default value.
  • This interface file is not stored in the memcg root directory.

Configuration examples

  1. Create a test file.
    mkdir /sys/fs/cgroup/memory/test/
  2. Specify the value of memory.limit_in_bytes.
    In this example, the value is 1 GiB. 
    echo 1G > /sys/fs/cgroup/memory/test/memory.limit_in_bytes
  3. Configure memory.wmark_ratio.
    In this example, the memcg memory watermark where asynchronous reclaim starts is set to 95% of the memcg memory limit. 
    echo 95 > /sys/fs/cgroup/memory/test/memory.wmark_ratio
  4. Run the cat /sys/fs/cgroup/memory/test/memory.wmark_scale_factor command to query the value of memory.wmark_scale_factor.
    The default value is 0.50% of the memcg memory limit. 
    50
  5. Query the values of memory.wmark_high and memory.wmark_low.
    If the following values are displayed for memory.wmark_high and memory.wmark_low in the command outputs, the configurations are correct.
    • If the following value is displayed for memory.wmark_high in the cat /sys/fs/cgroup/memory/test/memory.wmark_high command output: 
      1020051456
    • If the following value is displayed for memory.wmark_low in the cat /sys/fs/cgroup/memory/test/memory.wmark_low command output: 
      1014685696