Skip to content

Latest commit

 

History

History
72 lines (42 loc) · 2.75 KB

UPGRADE.md

File metadata and controls

72 lines (42 loc) · 2.75 KB
Raw

Upgrade notes

If you're using stolon as back-end for your LINSTOR installation you can easily perform the backup of your database:

kubectl exec -n linstor sts/linstor-db-stolon-keeper -- \
  sh -c 'PGPASSWORD=$(cat $STKEEPER_PG_SU_PASSWORDFILE) pg_dump -c -h linstor-db-stolon-proxy -U stolon linstor | gzip' \
  > linstor-backup.sql.gz

It always recommended to perform the backup before each LINSTOR upgrade.

Also if you were using helm template to perform the installation as described in README.md for v1.1.2, I would suggest you switch to Helmv3, however helm template method should also work fine, we're using it with qbec.

Anyway you can perform upgrade by simple replacing resources in your Kubernetes cluster thus

Upgrading helm repo index

helm repo update

Upgrading stolon

Helm way:

helm upgrade linstor-db kvaps/stolon --namespace linstor -f linstor-db.yaml

Templated manifests:

  • (optional) Remove all stolon resources except generated ones.
    The generated resources (like PVC's and stolon-cluster-linstor-db-stolon configmap), should remain in the cluster even after you remove mentioned statefulsets for them.

  • Install new resources with the same names, they should start using old PVCs and linstor-db-stolon configmap to reload cluster state. Specify --set job.autoCreateCluster=false option for stolon chart.

Upgrading LINSTOR

Helm way:

helm upgrade linstor kvaps/linstor --version 1.14.0 --namespace linstor -f linstor.yaml

Templated manifests:

  • (optional) Remove all LINSTOR resources, it does not store any state in the Kubernetes cluster, so you can do that without fear. The LINSTOR state is stored only in database.

  • Create new LINSTOR resources. check the controller log, it should perform the schema migration for the database.

If you're upgrading from old version you can see your nodes in Offline state, that's because latest version enables mutual ssl authentification for the linstor-satellites.

You can easily fix that by executing this command in your linstor-controller pod:

linstor n l | awk '/(PLAIN)/ { print "linstor n i m -p 3367 --communication-type SSL " $2 " default" }' | sh -ex

v1.9.0 release also introduce shorter release name: linstor- instead of linstor-linstor-, this change shouldn't break anything, however it will regenerate SSL certificates. If you are using LINSTOR API externally, you might need to update the client certificates or keep the old release name prefix by specifying --set fullnameOverride=linstor-linstor option.

See #18 for more details.