The following steps require that Lightwave has been configured to use data volumes. (See STS-CONTAINER-DEPLOYMENT.md)
For multiple replicas, backup/upgrade/rollback can be performed on individual nodes one at a time, only temporarily bringing down the node currently being operated on. In the case of an upgrade failure, only the failed node will need to be rolled back. Having nodes with different versions in a domain is supported.
Tag images
New images may have same name as the old image. To keep them organized, tag the image before loading a new image:
# docker tag vmware/lightwave-sts vmware/lightwave-sts-old
Before performing upgrade, backup all data for use in rollback in the case of upgrade failure.
# mkdir backup-6-6-0
This container will map a volume to a backup directory on the host, and can be used for backup and restore operations.
Create backup container:
# docker run -d --name backup --volumes-from lw_data -v $(pwd)/backup-6-6-0:/backup vmware/lightwave-sts
Install tar
For upgrading from lightwave 1.0 beta or rc, tar must be installed
# docker exec backup tdnf install tar
Create tars of all data volumes
# docker exec backup tar cvfP /backup/lib-vmware.tar /var/lib/vmware
# docker exec backup tar cvfP /backup/lib-likewise.tar /var/lib/likewise
# docker exec backup tar cvfP /backup/etc-likewise.tar /etc/likewise/
# docker exec backup tar cvfP /backup/etc-vmware-sso.tar /etc/vmware-sso
Remove backup container
# docker stop <backup container id>
# docker rm <backup container id>
Stop running Lightwave container
# docker stop <lw container name>
Create/edit config file so that domain, deployment and vmdir password in the lightwave-server config
# cat /var/lib/vmware/config/lightwave-server.cfg
deployment=standalone
domain=vsphere.local <<<<
admin=Administrator <<<<
password=Admin!23
first-instance=true
site-name=Default-first-site
hostname=10.118.97.160
Load new container:
# docker load < vmware-lightwave-sts.tar
Deploy a container with the new image using same data volumes as previous installation
Giving the container a distinguishing name may help with managing the many containers that may get used during the upgrade process.
During initialization, the container will detect if upgrade logic is to executed, put the directory into non-replication state and perform any data patching needed.
Example:
# docker run -d --name lightwave-1-1-0 --privileged --net=host -v /sys/fs/cgroup:/sys/fs/cgroup:ro -v /var/lib/vmware/config:/var/lib/vmware/config --volumes-from lw_data_container vmware/lightwave-sts
a689a33d718ff41692d230f4c39b3422e759f68d764c0a4a0638aca5af9af80f
If upgrade is successful, the directory will be taken out of non-replication state.
Check journalctl for upgrade completion:
# docker exec lightwave-1-1-0 journalctl | grep configure-lightwave-server
<snip>
Sep 10 00:12:10 photon-ga.eng.vmware.com configure-lightwave-server[64]: Running vdcupgrade
Sep 10 00:12:10 photon-ga.eng.vmware.com configure-lightwave-server[64]: Directory upgrade success.
Sep 10 00:12:10 photon-ga.eng.vmware.com configure-lightwave-server[64]: Upgrade complete.
Remove lightwave config file
This file contains administrator credentials and should be deleted when finished with upgrade.
# rm /var/lib/vmware/config/lightwave-server.cfg
If journalctl indicates an upgrade failure, a rollback may be performed to revert Lightwave container back to known good state.
Create a new 'rollback' container
As with backup, this container will map a volume to the backup directory on the host.
Note: If a data volumes container was created during installation, the failed container can be killed at this point, and the data container can provide the data volumes. Otherwise, the running failed container can provide the volumes since it is in a non-replication state.
Create rollback container using volumes from the container to be restored:
# docker run -d --name rollback --volumes-from lw_data -v $(pwd)/backup-6-6-0:/backup vmware/lightwave-sts
Extract tars of all data volumes
# docker exec rollback tar xvfP backup/lib-vmware.tar
# docker exec rollback tar xvfP backup/lib-likewise.tar
# docker exec rollback tar xvfP backup/etc-likewise.tar
# docker exec rollback tar xvfP backup/etc-vmware-sso.tar
Remove rollback container
# docker stop <rollback container id>
# docker rm <rollback container id>
If the failed upgrade container has not been stopped, do so before starting a container using the old image.
Restart old version container or use docker run to start a new instance of the old Lightwave container using the restored data:
# docker run -d --name vmsts-restored --privileged --net=host -v /sys/fs/cgroup:/sys/fs/cgroup:ro --volumes-from <backup image id> <original LW image id>