IBM Spectrum Scale installation toolkit - upgrade process flow

The upgrade procedure using the installation toolkit comprises two phases. In the first phase, all the non-protocol nodes are upgraded. In the second phase, all the protocol nodes are upgraded. The second phase is executed in two stages. NFS, object, performance monitoring, and GUI functions are upgraded in the first stage of phase 2. SMB and CES are upgraded in the second stage of phase 2.

Phase 1 - upgrade non-protocol nodes

In this phase, all non-protocol nodes are upgraded one node at a time. The first step is to check the cluster health. After this, the upgrade process starts on the first non-protocol node.

Check the cluster health.

Perform following steps on each non-protocol node, one node at a time.

Unmount file systems and shut down GPFS™.
Attention: At this point, I/O from this node stops.
Upgrade core GPFS packages.
Build GPFS portability layer (GPL).
Start up GPFS.
Perform node health check.
Mount the file systems.
Attention: At this point, I/O from this node resumes.

These steps are repeated on the remaining non-protocol nodes, one node at a time. After all of the non-protocol nodes are upgraded, phase 1 is complete.

Phase 2 - upgrade protocol nodes (stage 1)

In this stage of phase 2, all protocol nodes are upgraded one node at a time.

Check the node health.

Suspend CES on the node.

Attention: At this point, I/O from this node stops.

Upgrade core GPFS on the node. This procedure is same as phase 1.

Stop object services on all protocol nodes in the cluster.

Attention: At this point, object I/O stops in the whole cluster.

Upgrade object package.

Upgrade NFS, performance monitoring, and GUI as follows:

Stop the component.
Upgrade the component.
Start the component.

Resume CES on the node.

Attention: At this point, I/O from this node resumes.

Check the node health.

Steps 2 to 8 are repeated on all protocol nodes, one node at a time.

After the preceding steps are repeated on all protocol nodes, do the following steps.

Stop CES services.
Perform object version sync.
Start object services on all protocol nodes.
Attention: At this point, object I/O resumes in the whole cluster.

This concludes the first stage of phase 2.

Phase 2 - upgrade protocol nodes (stage 2)

In this stage of phase 2, all the protocol nodes are divided into two halves and each half is upgraded one by one as follows.

Suspend CES on the first half of nodes.

Stop SMB on the 1st half of nodes.

Attention: At this point, SMB I/O on this half of nodes stops and CES IPs on these nodes are migrated to the nodes in the other half.

Upgrade SMB and CES packages on the 1st half of nodes.

Suspend CES on the 2nd half of nodes.

Attention: At this point, SMB I/O stops on the whole cluster for a brief period.

Stop SMB on the 2nd half of nodes.

Start SMB on the 1st half of nodes.

Resume CES on the 1st half of nodes.

Attention: At this point, SMB I/O resumes on this half of nodes.

Upgrade SMB and CES packages on the 2nd half of nodes.

Start SMB on the 2nd half of nodes.

Resume CES on the 2nd half of nodes.

Attention: At this point, SMB I/O resumes in the whole cluster.

Check the cluster health.

After these steps are done, the installation toolkit automatically runs upgrade post checks. You can manually run the upgrade post checks by issuing the ./spectrumscale upgrade --postchecks command.

You must complete the upgrade to the new code level to take advantage of the new functionality. For more information, see Completing the upgrade to a new level of IBM Spectrum Scale.