Amazon ParallelCluster reserved users and groups
Amazon ParallelCluster creates a set of reserved Linux users and groups on all nodes
to run its internal daemons and services, such as the Slurm scheduler and
MUNGE authentication. These accounts are assigned
UIDs and GIDs from a reserved block, 400–405. Amazon ParallelCluster assigns each
account by name during node bootstrap, and relies on these names,
UIDs, and GIDs remaining stable and consistent across every node in the cluster.
Warning
Do not change the names, UIDs, GIDs, or ownership of the reserved users and groups listed on this page. Changing these values can cause cluster create, update, and delete (CRUD) operations and image build operations to fail. It can also break the Amazon ParallelCluster management daemons at run time.
The following table lists the users and groups that Amazon ParallelCluster creates and their default
UIDs and GIDs. Each user has a matching primary group with the same name and ID, except for
pcluster-slurm-share, which is a group with no associated user.
| Name | Default UID | Default GID | Purpose |
|---|---|---|---|
|
|
|
Amazon ParallelCluster administrative and service user, and its matching primary
group. Runs the management daemons (for example |
|
|
|
User and matching primary group for the Slurm scheduler daemons
( |
|
|
|
User and matching primary group for the MUNGE authentication daemon, which Slurm uses to authenticate communication between nodes. |
|
|
|
User and matching primary group for the Amazon DCV external authenticator. |
|
Not applicable |
|
Shared group used by the Slurm resume program. It has no associated user; the
|
Note
If you create additional users or groups of your own, we recommend that you assign them UIDs and GIDs outside the range that Amazon ParallelCluster reserves, to avoid conflicts with these accounts.