no quorum

[arrcy]

hi community, i have currently setup a cluster with 2 nodes. today i ran into a issue because i had to reboot one of the nodes for "maintenance". After making changes and booting the macine back up. i ran into no quorum

now i DO understand that

pvecm expected 1

temporary fixes the no quorum issue but here is MY problem

the node that i was doing maintenance on has a pcie network card passthourgh into a pfsense vm. wich ofc results in no quorum after boot since it needs quorum to start the vm's. but because the vm cant start there is no network availability. so my dumb idear was to just remove the cluster wich ended up screwing my time

i ended up running the following script that i found on https://gist.github.com/ianchen06/73acc392c72d6680099b7efac1351f56

echo -e "stopping Services:\npvestatd"
    systemctl stop pvestatd.service
    echo "pvedaemon"
    systemctl stop pvedaemon.service
    echo "pve-cluster.service"
    systemctl stop pve-cluster.service
    echo "pve-corosync"
    systemctl stop corosync
    echo "pve-cluster"
    systemctl stop pve-cluster
    echo "deleting data from db..."
    #echo "select * from tree where name = 'corosync.conf';"| sqlite3 /var/lib/pve-cluster/config.db
    echo "delete from tree where name = 'corosync.conf';"| sqlite3 /var/lib/pve-cluster/config.db
    echo "select * from tree where name = 'corosync.conf';"| sqlite3 /var/lib/pve-cluster/config.db
    #Remove directories
    pmxcfs -l
    rm -f /var/lib/pve-cluster/.pmxcfs.lockfile
    rm -f /etc/pve/corosync.conf
    rm -f /etc/corosync/*
    rm -f /var/lib/corosync/*
    rm -rf /etc/pve/nodes/$1/*
    echo "Staring services ..."
    echo "pvestatd"
    systemctl start pvestatd.service
    echo "pvedaemon"
    systemctl start pvedaemon.service
    echo "pve-cluster.service"
    systemctl start pve-cluster.service
    echo "pve-corosync"
    systemctl start corosync
    echo "pve-cluster"
    systemctl restart pve-cluster
else
    echo "Deleting node name missing or not existing. EXIT!"
    exit
fi

(^dont do this^)
without checking what was in the script i soon realised i made a verry horrible mistake. not having backups of configs i started coppying raw images and finding configs that were not effected. now 8 hours later i kinda got things working with some data loss. now you might think i dont want to make a cluster. but i dont give up so here is my question.

how can i make it that the node starts the vm's even tho there is no quorum and i am not typing in pvecm expected 1 through ssh/etc


thx for your time and as feature request i would like to have a unbind/remove all nodes from cluster/ thing in the web ui

 

[fabian]

how can i make it that the node starts the vm's even tho there is no quorum and i am not typing in pvecm expected 1 through ssh/etc

you can't. the whole point of the quorum checks is to not do dangerous stuff like this. you can override it temporarily (e.g. via 'pvecm expected') if you did all the necessary checks yourself, but then it's your fault if stuff goes wrong

 

[arrcy]

you can't. the whole point of the quorum checks is to not do dangerous stuff like this. you can override it temporarily (e.g. via 'pvecm expected') if you did all the necessary checks yourself, but then it's your fault if stuff goes wrong

But then again the machine wont start the routing vm, thus having no connection with the other machine. So you are teling me there is nothing i can do to start the vm on boot. maby having a pvecm expected 1 in an script.sh that starts on boot ( cron/rc.local
Or something else that is not recomended.

 

[fabian]

usually you'd not route your cluster traffic through a local VM, as obviously that creates a bootstrapping problem even on clusters with 3+ nodes..

 

[arrcy]

usually you'd not route your cluster traffic through a local VM, as obviously that creates a bootstrapping problem even on clusters with 3+ nodes..

I understand its not the best use case, but it does work quite well when things are running. besides i dont want to go out buying new machines because the setup i have does "work" but has some flaws. then again would it work to have some script running when the machine starts or is this gonne make more problems. if its too big of an issue i most likely just end up managing the nodes seperatly even tho it does defeat the idear of having "free" cluster feature with proxmox.

 

[arrcy]

I understand its not the best use case, but it does work quite well when things are running. besides i dont want to go out buying new machines because the setup i have does "work" but has some flaws. then again would it work to have some script running when the machine starts or is this gonne make more problems. if its too big of an issue i most likely just end up managing the nodes seperatly even tho it does defeat the idear of having "free" cluster feature with proxmox.

So my question was
would it work to have some script running when the machine starts or is this gonne make more problems

I have been managing the machines seperatly for some time now. And would still like to do something about it. Therefore i would like to know if it is worth the problems or not

 

[fabian]

yes, in this case not clustering or not routing your cluster network through the VM are the only sane options.