`lxc list` could be optimized

[simondeziel]

Please confirm

• I have searched existing issues to check if an issue already exists for the bug I encountered.

Distribution

Ubuntu

Distribution version

24.04

Output of "snap list --all lxd core20 core22 core24 snapd"

`latest/edge`

Output of "lxc info" or system info if it fails

n/a

Issue description

Listing instance information that comes from the DB is quick and optimal. Listing just the instance names with lxc list -f csv -c n will show like that in lxc monitor --pretty:

time="2025-10-08T15:28:14-04:00" level=debug msg="Handling API request" ip=@ method=GET protocol=unix url=/1.0 username=sdeziel
time="2025-10-08T15:28:14-04:00" level=debug msg="Handling API request" ip=@ method=GET protocol=unix url="/1.0/instances?filter=&recursion=1" username=sdeziel

Listing any column that requires to get state information not available int the DB seems to be much less optimal. When asking for the instance's PID (-c p) or instance's process count (-c N), all the state information is requested from every existing instance.

Here with the following instances:

$ lxc list -f csv -c n
ansible
ansible01
autopkg
juju
noble-builder
pylxd
v1
wine-games

A lxc list -f csv -c p would look like this in lxc monitor --pretty:

time="2025-10-08T15:37:43-04:00" level=debug msg="Handling API request" ip=@ method=GET protocol=unix url=/1.0 username=sdeziel
time="2025-10-08T15:37:43-04:00" level=debug msg="Handling API request" ip=@ method=GET protocol=unix url="/1.0/instances?filter=&recursion=2" username=sdeziel
time="2025-10-08T15:37:43-04:00" level=debug msg="Sending request to LXD" etag= method=GET url="https://custom.socket/1.0"
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage started" driver=zfs instance=autopkg pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage started" driver=zfs instance=ansible pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage started" driver=zfs instance=ansible01 pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="Sending request to LXD" etag= method=GET url="https://custom.socket/1.0/state"
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage finished" driver=zfs instance=ansible01 pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage finished" driver=zfs instance=ansible pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage finished" driver=zfs instance=autopkg pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage started" driver=zfs instance=v1 pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage started" driver=zfs instance=juju pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage started" driver=zfs instance=noble-builder pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage started" driver=zfs instance=pylxd pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage finished" driver=zfs instance=v1 pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage finished" driver=zfs instance=noble-builder pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage finished" driver=zfs instance=juju pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage finished" driver=zfs instance=pylxd pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage started" driver=zfs instance=wine-games pool=default project=default
time="2025-10-08T15:37:43-04:00" level=debug msg="GetInstanceUsage finished" driver=zfs instance=wine-games pool=default project=default

Those GetInstanceUsage are about getting disk usage information which is pointless and potentially slow, especially on ceph and/or when there are many instances.

Steps to reproduce

In a terminal, run:

lxc monitor --pretty | grep -E ' (instance|url)='

In another terminal, run one of:

lxc list -f csv -c n  # name check, fast
lxc list -f csv -c p  # PID check, slow
lxc list -f csv -c N # proc count, slow

Information to attach

• Any relevant kernel output (dmesg)
• Instance log (lxc info NAME --show-log)
• Instance configuration (lxc config show NAME --expanded)
• Main daemon log (at /var/log/lxd/lxd.log or /var/snap/lxd/common/lxd/logs/lxd.log)
• Output of the client with --debug
• Output of the daemon with --debug (or use lxc monitor while reproducing the issue)

[simondeziel]

I think this is related to this known race in the test suite:

# XXX: `[ "$(lxc list -f csv -c s "${name}")" = "RUNNING" ]` is too fast and leads to races with exec/delete
lxc list --fast "${name}" | grep -wF RUNNING