ivampiresp/longhorn
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Update upgrade-from-v0.6.2-to-v0.7.0.md

Browse Source 
Signed-off-by: Sheng Yang <sheng.yang@rancher.com>
This commit is contained in:
Sheng Yang 2019-11-14 17:33:33 -08:00 committed by GitHub
parent dcd1ae1eb2
commit 64c913d6c9
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 11 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
docs/upgrade-from-v0.6.2-to-v0.7.0.md
View File

@ -2,14 +2,11 @@
The users need to follow this guide to upgrade from v0.6.2 to v0.7.0. The users need to follow this guide to upgrade from v0.6.2 to v0.7.0.
## Warning
Longhorn v0.7.0 breaks the regular rollback and some risky operations are needed as the workaround.
Please check the rollback part if you may want to do rollback later before upgrading to v0.7.0.
## Preparation ## Preparation
1. Make backups for all the volumes. 1. Make backups for all the volumes.
1. Stop the workload using the volumes. 1. Stop the workload using the volumes.
1. Live upgrade is not supported from v0.6.2 to v0.7.0
## Upgrade ## Upgrade
### Use Rancher App ### Use Rancher App
@ -79,8 +76,7 @@ kubectl delete -f https://raw.githubusercontent.com/longhorn/longhorn/v0.7.0/exa
### Error: `kind CustomResourceDefinition with the name "xxx" already exists in the cluster and wasn't defined in the previous release...` ### Error: `kind CustomResourceDefinition with the name "xxx" already exists in the cluster and wasn't defined in the previous release...`
- This is [a Helm bug](https://github.com/helm/helm/issues/6031). - This is [a Helm bug](https://github.com/helm/helm/issues/6031).
- Please make sure that you have not deleted the old Longhorn CRDs via the command `curl -s https://raw.githubusercontent.com/longhorn/longhorn-manager/master/hack/cleancrds.sh | bash -s v062` or Longhorn uninstaller before executing the following command. - Please make sure that you have not deleted the old Longhorn CRDs via the command `curl -s https://raw.githubusercontent.com/longhorn/longhorn-manager/master/hack/cleancrds.sh | bash -s v062` or executed Longhorn uninstaller before executing the following command. Otherwise you MAY LOSE all the data stored in the Longhorn system.
Otherwise you may lose all the data stored in the Longhorn system.
1. Clean up the leftover: 1. Clean up the leftover:
``` ```
@ -91,8 +87,14 @@ curl -s https://raw.githubusercontent.com/longhorn/longhorn-manager/master/hack/
2. Re-click the `Upgrade` button in the Rancher UI. 2. Re-click the `Upgrade` button in the Rancher UI.
## Rollback ## Rollback
Since we upgrade the CSI framework from v0.4.2 to v1.1.0 in this release, rolling back from Longhorn v0.7.0 to v0.6.2 or lower means backward upgrade for the CSI plugin. Since we upgrade the CSI framework from v0.4.2 to v1.1.0 in this release, rolling back from Longhorn v0.7.0 to v0.6.2 or lower means backward upgrade for the CSI plugin.
But Kubernetes does not support the CSI backward upgrade. **Hence restarting kubelet is unavoidable. Please be careful about the risky operation!** But Kubernetes does not support the CSI backward upgrade. **Hence restarting kubelet is unavoidable. Please be careful, check the conditions beforehand and follow the instruction exactly.**
Prerequisite:
* To rollback from v0.7.0 installation, you must haven't executed [the post upgrade steps](#post-upgrade).
Steps to roll back:
1. Clean up the components introduced by Longhorn v0.7.0 upgrade 1. Clean up the components introduced by Longhorn v0.7.0 upgrade
``` ```
@ -100,7 +102,8 @@ kubectl delete -f https://raw.githubusercontent.com/longhorn/longhorn/v0.7.0/exa
curl -s https://raw.githubusercontent.com/longhorn/longhorn-manager/master/hack/cleancrds.sh | bash -s v070 curl -s https://raw.githubusercontent.com/longhorn/longhorn-manager/master/hack/cleancrds.sh | bash -s v070
``` ```
2. Restart the Kubelet container on all nodes running Longhorn system 2. Restart the Kubelet container on all nodes or restart all the nodes. This step WILL DISRUPT all the workloads in the system.
Connect to the node then run Connect to the node then run
``` ```
docker restart kubelet docker restart kubelet