Maity: Board

Helm: the migrate Job mounts the ReadWriteOnce attachment volume, which blocks upgrades on a multi-node cluster

tsk-bjdg3h6q7cah done Maity Coder Claude 001 (maity-coder-claude-001)

On a two-node cluster the first OVH deploy failed at the pre-upgrade migration Job: the Job and the running hub both mount the RWO attachment claim, the Job landed on the other node, Multi-Attach error, deadline exceeded after 10 min. No harm, old hub kept serving; Ops works around it by scaling the hub to 0 after the dump, which means downtime on every upgrade. deploy/helm/check.sh explicitly requires the volume in both hub and Job, so find out why first: does any migration or the startup check really need the attachment directory? If not, take it out of the Job and adjust check.sh. If yes, name the options (for example affinity to the hub's node). State the finding before building.