plcontainer reports: cgroup-parent for systemd cgroup should be a valid slice named as "xxx.slice"

Products

• VMware Tanzu Greenplum
• Greenplum
• VMware Tanzu Data Suite
• VMware Tanzu Data Suite

Issue/Introduction

When running a plcontainer function with resource groups, (See PL/Container Resource Management), it reports and error for the cgroup-parent:

postgres=# select dummyPython();
ERROR:  plcontainer: backend create error (containers.c:405)
DETAIL:  Failed to create container, return code: 400, detail: {"message":"cgroup-parent for systemd cgroup should be a valid slice named as \"xxx.slice\""}
CONTEXT:  PLContainer function "dummypython"
postgres=#

Cause

The issue is caused by the way the docker is started in the systemd/systemctl. Check with "docker info":

docker info

Client:
 Version:    27.3.1
 Context:    default
 Debug Mode: false
  ...
  Using metacopy: true
  Native Overlay Diff: false
  userxattr: false
 Logging Driver: json-file
 Cgroup Driver: systemd           <------------ NOTE HERE
 Cgroup Version: 2
 Plugins:
  Volume: local
  Network: bridge host ipvlan macvlan null overlay
  ...
  127.0.0.0/8
 Live Restore Enabled: false 

Resolution

To resolve the issue, use the following command to update docker's configuration.

sudo systemctl edit docker.service

and set the cgroup driver to cgroupfs by appending '--exec-opt native.groupdriver=cgroupfs' to the docker start command.

[Service]
...
ExecStart=/path/to/dockerd ... --exec-opt native.cgroupdriver=cgroupfs
...

Then restart docker by

sudo systemctl restart docker

The re-run the plcontainer function. The container process should be controlled by gpdb's resource group under '/sys/fs/cgroup/gpdb.service'.