[SOLVED] Proxmox 5.0 user start up scripts.

[tomtom13]

Hi, this is kinda trivial question, but I found usual scripts that should run at the end of level 6 init process but since proxmox and the rest of the world is mowing towards systemd it's all a bit messy to me and it seems that usual suspects don't get run as last ...

Also I would like the run the script when
- proxmox is considered "up and running"
- ceph is considered up and running (at least all systemd starting processes)

this is because I'm intending to mount cephfs to a specific folder and doing it from fstab is just to early.

 

[wolfgang]

Hi,

you can use crontab with @reboot

see man 5 crontab

or you can use systemd.timer
with OnStartupSec=

see
https://wiki.archlinux.org/index.php/Systemd/Timers

 

[tomtom13]

Thing is that @reboot will run at the point where console is initiated (level 6 when login prompt shows up) - while ceph services are actually still starting up. I've added ceph-mds@0 and you can see that one still being in state starting up (along with whole load of ceph-osd@0 services)

I'll look into timers, still It would be nice if there was a default user script in proxmox which i being run when proxmox is fully up but before any VM will run and a script that will be executed afterwards.

Any other advice appreciated.

 

[fabian]

Thing is that @reboot will run at the point where console is initiated (level 6 when login prompt shows up) - while ceph services are actually still starting up. I've added ceph-mds@0 and you can see that one still being in state starting up (along with whole load of ceph-osd@0 services)

I'll look into timers, still It would be nice if there was a default user script in proxmox which i being run when proxmox is fully up but before any VM will run and a script that will be executed afterwards.

Any other advice appreciated.

the on boot guests are started by pveproxy (non HA) and pve-ha-lrm (HA), so just order yourself between those and whatever you need (pve-cluster for pmxcfs, storage units for various storages, ...).

 

[tomtom13]

So, thanks for all the replies - it got me going the right way.

I anybody ever looks for stuff like that, here is my dirty script:

/etc/systemd/system/multi-user.target.wants/mount-cephfs.service

[Unit]
Description=Mount cephsFS info /cephfs
ConditionPathExists=/cephfs
Wants=pve-cluster.service
Wants=pvedaemon.service
Wants=ssh.service
Wants=remote-fs.target
Wants=pveproxy.service
Wants=ceph-mds@0.service
After=pve-cluster.service
After=pvedaemon.service
After=ssh.service
After=remote-fs.target
After=ceph.service
After=sheepdog.service
After=glusterd.service
After=ceph-mds@0.service
Before=pveproxy.service

[Service]
ExecStartPre=/bin/sleep 30
ExecStart=/bin/mount /cephfs
ExecStop=/bin/umount /cephfs
Type=oneshot
RemainAfterExit=yes

[Install]
WantedBy=multi-user.target

Just "FYI"
this script is not perfect, and if you woke up to a cluster that is broken and it's in warn / error and can't mount this script will fail anyway.