This bug was initially created as a copy of Bug #1828758

I am copying this bug because: 

Also applies.


Description of problem:

On large systems with many users connecting to it (e.g. ~1000 users, but can be reproduced with less users, depending on how many sessions they use), a simple "systemctl list-unit-files" query freezes systemd for an enormous amount of time (>= 50 seconds), preventing regular users to use the system:
- they cannot connect to it in less than 25 seconds
- no event is handled by systemd (clean ups, etc)

The reason behind this is systemd not off-loading the "ListUnitFiles" query which is heavily I/O consuming (it reads *all* unit configuration files).

Version-Release number of selected component (if applicable):

systemd-219 but also systemd-239 (RHEL8) and later


How reproducible:

Always


Steps to Reproduce:
1. Create 1000 users in a VM with 8GB of memory / 4 CPUs

  # for i in $(seq 1 1000); do user="user$(printf "%04d" $i)"; echo $user; useradd $user; echo "user" | passwd --stdin $user; done

2. Copy root's key to user0001's authorized_keys

  # ssh-copy-id user0001@localhost
  Password: "user"

3. Create all other authorized_keys files

  # for i in $(seq 2 1000); do user="user$(printf "%04d" $i)"; cp -r user0001/.ssh $user; chown -R $user:$user $user/.ssh; done

4. Spawn 1000 ssh sessions

  # for i in $(seq 1 1000); do user="user$(printf "%04d" $i)"; echo "$user"; ssh -f $user@localhost 'sleep 3600'; done

5. Execute "list-unit-files"

  # systemctl list-unit-files --no-pager
  Failed to list unit files: Connection timed out


Actual results:

systemd not functional ~50 seconds!

Expected results:

systemd still doing its usual job


Additional info:

The "ListUnitFiles" is only an example, we suppose that there are other operations (DBus or others) that affect systemd as well, but so far we weren't able to find those yet.