Skip to content

Upgrades

An upgrade replaces the binary and restarts the service. On its first start the new version brings the data up to date by itself: database migrations and, when a release changes it, the on-disk layout. Read the changelog of every version you skip before upgrading.

Before you start

Download a fresh encryption recovery bundle (Settings → Security) and check that your last backup is recent.

Single node

wget https://github.com/MaxIOFS/MaxIOFS/releases/download/v1.10.0/maxiofs_v1.10.0_amd64.deb
sudo dpkg -i maxiofs_v1.10.0_amd64.deb

The package keeps /etc/maxiofs/config.yaml and restarts the service if it was running.

sudo rpm -U https://github.com/MaxIOFS/MaxIOFS/releases/download/v1.10.0/maxiofs-1.10.0-1.x86_64.rpm
sudo systemctl restart maxiofs

docker pull maxiofs/maxiofs:latest
docker stop maxiofs && docker rm maxiofs
docker run -d --name maxiofs ...   # the same command as before

With Compose: docker compose pull && docker compose up -d.

wget -O maxiofs.new https://github.com/MaxIOFS/MaxIOFS/releases/download/v1.10.0/maxiofs-linux-amd64-v1.10.0
sudo install -m 755 maxiofs.new /usr/local/bin/maxiofs
sudo systemctl restart maxiofs

Then check the version in the console (About, or the bottom of the side menu) and watch the log for errors:

sudo journalctl -u maxiofs -n 100

Maintenance mode

To keep clients from writing while you upgrade, turn on maintenance mode in Settings → System before, and off after. While it is on:

  • S3 reads (GET, HEAD) work; writes and deletes are refused.
  • The console shows a banner to everyone and refuses changes, except settings and sign-in.

A short upgrade does not need it: a restart takes seconds, and S3 clients retry.

Cluster

Upgrade one node at a time, so the others keep serving:

  1. Take the node out of the load balancer.
  2. Upgrade it as above and wait until Cluster → Nodes shows it healthy.
  3. Put it back in the load balancer.
  4. Repeat with the next node.

Same version everywhere

Upgrade every node soon after the first. Turn on parity or create a storage class only once all nodes run the same version.

Downgrades

A newer version can change the database or the layout in a way an older one cannot read. Going back means restoring the backup taken before the upgrade.