PCF v2.6 Breaking Changes
Page last updated:
This topic describes the breaking changes you need to be aware of when upgrading to Pivotal Cloud Foundry (PCF) v2.6. For more information about important preparation steps you must follow before beginning an upgrade, see Upgrading Pivotal Cloud Foundry.
See the following Ops Manager breaking changes:
Operators cannot upgrade the platform from PAS v2.5.10 or later to PAS v2.6 until all app instances (AIs) are using cflinuxfs3. If you are running PAS v2.5.10 or a later v2.5 patch, you must migrate all AIs to cflinuxfs3.
During upgrade, you must check a box to verify that all AIs have been migrated. If you check the box and the AIs have not been migrated, the upgrade causes all AIs to fail.
recommends that you communicate the change of stack to your development teams and have them migrate to cflinuxfs3 before upgrading to PAS v2.6.
For more information about this change, see PAS Only Supports cflinuxfs3 in Pivotal Application Service v2.5 Release Notes.
You can only SSH onto the Ops Manager VM in a vSphere deployment with a private SSH key. You cannot SSH onto the Ops Manager VM with a password.
To avoid upgrade failure and errors when authenticating, add a public key to the OVF template for the Ops Manager VM. Then, use the private key to SSH onto the Ops Manager VM.
For more information about adding a public key to the OVF template, see Deploy Ops Manager in Deploying Ops Manager on vSphere.
Warning: You cannot upgrade successfully without adding a public key. If you do not add a key, Ops Manager shuts down automatically because it cannot find a key and may enter a reboot loop.
stemcell property for the
/api/v0/diagnostic_report endpoint is removed. Update any automation that uses the stemcell information in the
/api/v0/diagnostic_report endpoint to avoid errors.
Information about all stemcells for both staged and deployed products is listed under a new
added_products property for the
/api/v0/diagnostic_report endpoint. Stemcell information under the
added_products property includes the stemcell filename, OS, and version.
For more information about the
/api/v0/diagnostic_report endpoint, see Diagnostic Report in the Ops Manager API documentation.
/api/v0/staged/pending_changes endpoint, the
stemcell_version key from previous versions of the Ops Manager API is now the
version sub-key under the
Update any automation that uses the stemcell information in the
/api/v0/pending_changes endpoint to avoid errors.
For more information about the
/api/v0/staged/pending_changes endpoint, see View a list of pending changes in the Ops Manager API documentation.
Ops Manager v2.6 uses the latest generation of instances for AWS deployments. When you upgrade to Ops Manager v2.6, your VM types convert to the corresponding 5th generation instance types, such as
If you are using a custom instance type catalog, your VM types that are used and available do not change.
These instance types are supported by Xenial stemcells and Trusty stemcell 3586 and later. Stemcells earlier than Trusty stemcell 3586 fail to boot Ops Manager. To avoid this, ensure that you are using Trusty stemcell 3586 or later.
For more information about AWS instance types, see Amazon EC2 Instance Types in the AWS documentation.
stemcells_assignments endpoint in the Ops Manager API has been deprecated. The deprecated endpoint only returns a single stemcell for a product.
Instead, operators should use the
stemcell_associations endpoint, which returns information about all stemcells associated with a product.
For more information, see Stemcell Associations for Products in the Ops Manager API documentation.
See the following PAS breaking changes:
If your SSH load balancer is configured with an HTTP health check, remove the healthcheck before upgrading. This is the load balancer specified in the Diego Brain row of the Resource Config pane.
Note: Neither the Terraform templates for installing PAS nor the PAS installation instructions specify a HTTP healthcheck for the SSH Load balancer.
ssh_proxy job running on the Diego Brain VM no longer starts a HTTP health check server. This feature was previously added to allow GCP load balancers to health check the service. However, GCP load balancers no longer require an HTTP endpoint for health checks.
Additionally, making this healthcheck server inaccessible improves the security of the platform.
PAS v2.6 increases the default and minimum CPU core count from one to two for Gorouter VMs.
Verify that you have sufficient infrastructure resources allocated to the deployment before upgrading to PAS v2.6.
In PAS v2.6, the NFS-Legacy Mounter is removed. Migrating to the NFS volume service may be required if
nfs-legacy services are bound with a
gid set to
0 or root. NFS Volume service does not allow specifying
gid=0 as bind options.
For more information on how to migrate apps from
nfs, see Upgrade Preparation Checklist for PCF v2.6.
In PAS v2.6.8 and later, the buffer pool size for the internal MySQL database is changed from 2 GB to 50% of available memory.
You may need to choose a VM type with a larger amount of RAM to avoid Out of Memory (OOM) errors.
For example, in PAS v2.6.7 and earlier, if the internal MySQL VM is configured with 8 GB of RAM and 6 GB is currently used, then 2 GB is consumed by the buffer pool and 4 GB is consumed by everything else.
In this example, upgrading to PAS v2.6.8 causes OOM errors because the buffer pool consumes 4 GB, which leaves 0 GB of available RAM.
See the following PASW breaking changes:
The Use “cf” as deployment name in emitted metrics instead of unique name option is removed from the Adavanced Features pane of the PASW tile.
Windows stemcell v2019.2 is not compatible with Ops Manager v2.6. Before upgrading to Ops Manager v2.6, upgrade your Windows stemcell to v2019.3 or later.