Test for scalability of creating thin snapshots of a thin LV containing a rootfs

test-bare-snapshot-limit.sh

#!/bin/bash

NUM_LVS=200
mkdir -p test-mounts

loop_count=0
while true; do
    [ $loop_count -ge $NUM_LVS ] && break;
    echo "Creating lv $loop_count"
    start=$(date +%s)
    sudo lvcreate -kn -n "test-lv-$loop_count" -s /dev/mantest_vg/04aac4257341478b49c25d22cea8a6ce0489dc6c42d835367945e7596368a37f || { echo "Failure to create # $loop_count"; break;}

    # step 2:
    echo "Activating"
    sudo lvchange -ay "/dev/mantest_vg/test-lv-$loop_count"

    #step 3:
    echo "Mounting"
    mkdir -p "test-mounts/test-lv-$loop_count"
    sudo mount "/dev/mantest_vg/test-lv-$loop_count" "test-mounts/test-lv-$loop_count"

    # step 4:
    echo "Writing something"
    echo "$loop_count" > "test-mounts/test-lv-$loop_count/rootfs/home/ubuntu/$loop_count.txt"

    end=$(date +%s)
    elapsed=$(($end - $start))
    echo "Created lv $loop_count in $elapsed"
    loop_count=$(($loop_count + 1))
done

read -p "waiting" x

while true; do
    loop_count=$(($loop_count - 1))
    [ $loop_count -eq -1 ] && break;
    echo "Removing test-lv-$loop_count"
    sudo umount "test-mounts/test-lv-$loop_count"
    sudo lvremove -f "mantest_vg/test-lv-$loop_count" || break;
done

In my testing, this runs on a NUC, the mantest_vg volumegroup contains one PV, an external USB drive.
The volume /dev/mantest-vg/04aac4257341478b49c25d22cea8a6ce0489dc6c42d835367945e7596368a37f is an untarred LXD image on an ext4 fs.
The thin pool that the original volume is in is 1t in size.

This runs with no problems up to 200 containers (haven't bothered testing higher). Each container takes about 1-2 seconds with basically no deviations