Derived Parameters

TripleO can populate environment files with parameters which are derived from formulas. Such formulas can take introspected hardware data, workload type, or deployment type as input and return as output a system tuning to be applied via parameter overrides. TripleO supports this feature for both NFV (Network Function Virtualization) and HCI (Hyper-converged Infrastructure; nodes with collocated Ceph OSD and Nova Compute services) deployments.

Using derived parameters during a deployment

To have TripleO derive parameters during deployment, specify an alternative deployment plan containing directives which trigger either a Mistral workflow (prior to Victoria) or an Ansible playbook (in Victoria and newer) which derives the parameters.

A default deployment plan is created during deployment. This deployment plan may be overridden by passing the -p or --plan-environment-file option to the openstack overcloud deploy command. If the plan-environment-derived-params.yaml file, located in /usr/share/openstack-tripleo-heat-templates/plan-samples/, is specified this way, then the process to derive parameters, via a playbook or workflow, will be executed.

The process is able to determine which roles have which features. E.g. ComputeHCI roles need parameters derived from HCI formulas and ComputeOvsDpdk roles need parameters derived from NFV formulas. The deployment definition will indicate which hardware in the Ironic database may be used and the process will extract the relevant introspection data to use as input in the derivation. The output of the process will be saved in the deployment plan and applied as specific settings for different roles. If a role uses both NFV and HCI features, then the process will apply NFV tuning first and then limit the available HCI resources accordingly. E.g. if an NFV process is pinning an entire CPU, then that CPU shouldn’t be considered available for use by a Ceph OSD.

Parameters which are derived for HCI deployments

The derived parameters for HCI sets the NovaReservedHostMemory and NovaCPUAllocationRatio per role based on the amount and type of Ceph OSDs requested during deployment, the available hardware in Ironic, and the average Nova guest workload.

Deriving the parameters is useful because in an HCI deployment the Nova scheduler does not, by default, take into account the requirements of the Ceph OSD services which are collocated with the Nova Compute services. Thus, it’s possible for Compute resources needed by an OSD to be given instead to a Nova guest. To prevent this from happening, the NovaReservedHostMemory and NovaCPUAllocationRatio may be adjusted so that Nova reserves the resources that the Ceph OSDs need.

To perform well, each Ceph OSD requires 5 GB of memory and a certain amount of vCPU (if hyperthreading is enabled). The faster the storage medium the more vCPUs an OSD should use in order for the CPU resources to not become a performance bottle-neck. All of this is taken into account by the derived parameters for HCI.

The workload of the Nova guests may also be taken into account. The plan-environment-derived-params.yaml file contains the following:

hci_profile: default
hci_profile_config:
  default:
    average_guest_memory_size_in_mb: 0
    average_guest_cpu_utilization_percentage: 0
  many_small_vms:
    average_guest_memory_size_in_mb: 1024
    average_guest_cpu_utilization_percentage: 20
  few_large_vms:
    average_guest_memory_size_in_mb: 4096
    average_guest_cpu_utilization_percentage: 80
  nfv_default:
    average_guest_memory_size_in_mb: 8192
    average_guest_cpu_utilization_percentage: 90

The hci_profile_config describes the requirements for the average Nova guest. The provided profiles (many_small_vms, few_large_vms, and nfv_default) are not necessarily prescriptive but are examples that may be adjusted based on your average workload. For example the many_small_vms profile indicates that the average Nova guest uses a flavor with 1 GB of RAM and that it is anticipated that the guest will use on average 20% of it’s available CPU. The hci_profile parameter should be set to a profile describing the expected workload.

In Victoria and newer the default profile sets the average memory and CPU utilization to 0 because by default the workload is unknown. When the tripleo_derive_hci_parameters Ansible module is passed these values it sets the NovaReservedHostMemory to the number of OSDs requested multiplied by 5 (for 5 GBs of RAM per OSD) but is not able to take into account the memory overhead per guest for the hypervisor. It also does not set the NovaCPUAllocationRatio. Thus, passing an expected average workload will produce a more accurate set of derived HCI parameters. However, this default does allow for a simpler deployment where derived parameters may be used without having to specify a workload but the OSDs are protected from having their memory allocated to Nova guests.

Deriving HCI parameters before a deployment

The tripleo_derive_hci_parameters Ansible module may be run independently on the undercloud before deployment to generate a YAML file to pass to the openstack overcloud deploy command with the -e option. If this option is used it’s not necessary to derive HCI parameters during deployment. Using this option also allows the deployer to quickly see the values of the derived parameters.

Warning

This playbook computes HCI parameters without running OVS-DPDK and SRIOV roles and may result in incorrect values in case OVS-DPDK and SRIOV are enabled.

On the undercloud within /usr/share/ansible/tripleo-playbooks/ a simple playbook derive-local-hci-parameters.yml is available which calls the tripleo_derive_hci_parameters Ansible module. To use the playbook before deployment determine the Ironic node UUID which will correspond to the role being deployed. In the example below a server with the name ceph-2 has already been introspected and will be used later during deployment for servers in the ComputeHCI role. We will use this server’s Ironic UUID so that the playbook gets its introspection data:

[stack@undercloud ~]$ openstack baremetal node list | grep ceph-2
| ef4cbd49-3773-4db2-80da-4210a7c24047 | ceph-2       | None          | power off   | available | False       |
[stack@undercloud ~]$

Make a copy of the playbook in the stack users home directory and then modify it to set the four playbook variables as below:

[stack@undercloud ~]$ head derive-local-hci-parameters.yml
---
- name: Derive HCI parameters before deployment
  hosts: localhost
  gather_facts: false
  vars:
    # Set the following variables for your environment
    ironic_node_id: ef4cbd49-3773-4db2-80da-4210a7c24047
    role: ComputeHCI
    average_guest_cpu_utilization_percentage: 50
    average_guest_memory_size_in_mb: 8192
    heat_environment_input_file: /home/stack/ceph_overrides.yaml
[stack@undercloud ~]$

In the above example it is assumed the role ComputeHCI will use nodes with the same type of hardware which is set to the ironic_node_id and that the average guest will use 50% of its CPU and will use 8 GB of RAM. If the workload is unknown, remove these variables. The system tuning will not be as accurate but the Ansible module will at least set the NovaReservedHostMemory as a function of the number of OSDs.

The heat_environment_input_file must be set to the path of the Heat environment file which defines the OSDs.

Victoria or earlier

When ceph-ansible is used, in place of cephadm, this should be the file where the CephAnsibleDisksConfig parameter is set. This parameter is used to define which disks are used as Ceph OSDs and might look like the following if bluestore was being deployed on 4 NVMe SSDs:

parameter_defaults:
  CephAnsibleDisksConfig:
    osd_scenario: lvm
    osd_objectstore: bluestore
    osds_per_device: 4
    devices:
      - /dev/nvme0n1
      - /dev/nvme0n2
      - /dev/nvme0n3
      - /dev/nvme0n4

The derived parameters workflow would use the values above to determine the number of OSDs requested (e.g. 4 devices * 4 OSDs per device = 16) and the type of device based on the Ironic data (e.g. during introspection, ironic can determine if a storage device is rotational).

If cephadm is used, in place of ceph-ansible (for Wallaby and newer), then the heat_environment_input_file must be set to the path of the file where the CephHciOsdCount and CephHciOsdType parameters are set.

The CephHciOsdCount and CephHciOsdType exist because CephOsdSpec, as used by cephadm, might only specify a description of devices to be used as OSDs (e.g. “all devices”), and not a list of devices like CephAnsibleDisksConfig, setting the count directly is necessary in order to know how much CPU/RAM to reserve. Similarly, because a device path is not hard coded, we cannot look up that device in Ironic to determine its type. For information on the CephOsdSpec parameter see the Deploying Ceph with cephadm documentation.

CephHciOsdType is the type of data_device (not db_device) used for each OSD and must be one of hdd, ssd, or nvme. These are used by the Ansible module tripleo_derive_hci_parameters.

CephHciOsdCount is the number of expected Ceph OSDs per HCI node. If a server has eight HDD drives, then the parameters should be set like this:

parameter_defaults:
  CephHciOsdType: hdd
  CephHciOsdCount: 8

To fully utilize nvme devices for data (not metadata), multiple OSDs are required. If the CephOsdSpec parameter is used to set osds_per_device to 4, and there are four NVMe drives on a host (and no HDD drives), then the parameters should be set like this:

parameter_defaults:
  CephHciOsdType: nvme
  CephHciOsdCount: 16

After these values are set run the playbook:

[stack@undercloud ~]$ ansible-playbook derive-local-hci-parameters.yml
[WARNING]: provided hosts list is empty, only localhost is available. Note that the implicit
localhost does not match 'all'

PLAY [Derive HCI parameters before deployment] ***********************************************

TASK [Get baremetal inspection data] *********************************************************
ok: [localhost]

TASK [Get tripleo CephDisks environment parameters] *******************************************
ok: [localhost]

TASK [Derive HCI parameters] *****************************************************************
changed: [localhost]

TASK [Display steps on what to do next] ******************************************************
ok: [localhost] => {
    "msg": "You may deploy your overcloud using -e /home/stack/hci_result.yaml so that the role ComputeHCI has its Nova configuration tuned to reserve CPU and Memory for its collocated Ceph OSDs. For an explanation see /home/stack/hci_report.txt."
}

PLAY RECAP ***********************************************************************************
localhost                  : ok=4    changed=1    unreachable=0    failed=0    skipped=0    rescued=0    ignored=0

[stack@undercloud ~]$

The playbook will generate two files in the stack user’s home directory unless the new_heat_environment_output_file and report_path variables are modified. The file denoted by the first variable generated will be the derived parameters for the role specified. For example:

[stack@undercloud ~]$ cat /home/stack/hci_result.yaml
parameter_defaults:
  ComputeHCIParameters:
    NovaCPUAllocationRatio: 8.2
    NovaReservedHostMemory: 75000
[stack@undercloud ~]$

The above could be used during a deployment by running a command like openstack overcloud deploy ... -e /home/stack/hci_result.yaml. The hci_result.yaml should be appended near the end of the openstack overcloud deploy command so that the derived values take precedence.

The second file, defined by the report_path variable, will contain an explanation of how the parameters were derived and what relevant information was provided as input including the disks types as found in Ironic. It might look like the following:

[stack@undercloud ~]$ cat /home/stack/hci_report.txt
Derived Parameters results
 Inputs:
 - Total host RAM in GB: 256
 - Total host vCPUs: 56
 - Ceph OSDs per host: 10
 - Average guest memory size in GB: 2
 - Average guest CPU utilization: 10%

 Outputs:
 - number of guests allowed based on memory = 90
 - number of guest vCPUs allowed = 460
 - nova.conf reserved_host_memory = 75000 MB
 - nova.conf cpu_allocation_ratio = 8.214286

Compare "guest vCPUs allowed" to "guests allowed based on memory"
for actual guest count.

OSD type distribution:
  HDDs 10 | Non-NVMe SSDs 0 | NVMe SSDs 0
  vCPU to OSD ratio: 1
[stack@undercloud ~]$

Verifying that HCI derived parameters have been applied

If derived parameters were computed during deployment, then their parameter override outputs may be found in the deployment plan. Download the deployment plan for the stack, e.g. overcloud with a command like the following:

openstack overcloud plan export overcloud
tar xf overcloud.tar.gz

Locate the plan-environment.yaml file and check if it contains the the derived NovaCPUAllocationRatio and NovaReservedHostMemory, for example:

$ head -5 plan-environment.yaml
derived_parameters:
  ComputeHCIParameters:
    NovaCPUAllocationRatio: 8.2
    NovaReservedHostMemory: 75000
description: 'Default Deployment plan'
$

Regardless of if the parameters were derived before or during the deployment, they should be applied to the overcloud. The following example shows commands being executed on a node from the ComputeHCI role and where expected Nova settings were applied:

$ sudo podman exec -ti nova_compute /bin/bash
# egrep 'reserved_host_memory_mb|cpu_allocation_ratio' /etc/nova/nova.conf
reserved_host_memory_mb=75000
cpu_allocation_ratio=8.2
#

Migrating Away from Derived Paramters

If a stack update is run without the -p option (and it was used previously), then any derived parameters will no longer be applied to the overcloud. An example of this is the following process:

  1. Deploy an overcloud using -p plan-environment-derived-params.yaml to set a reserved_host_memory_mb

  2. Observe the reserved_host_memory_mb in /etc/nova/nova.conf

  3. Run a stack update without using -p plan-environment-derived-params.yaml

  4. Observe that the derived reserved_host_memory_mb will not be in the /etc/nova/nova.conf

Thus, to migrate away from using derived parameters, take note of what parameters and values were derived and then create a Heat enviornment file which sets them directly.

One way to take note what parameters and values were derived is to look search the output of the openstack overcloud deploy -p … comamnd for the string “derived_parameters”. For example:

"derived_parameters": {
    "ComputeHCIParameters": {
        "NovaReservedHostMemory": 15360
    }
},

If the above is not available, then it can be observed directly on the system which was deployed with derived parameters. For example:

[root@compute-hci-2 ~]# podman exec -ti nova_compute /bin/bash
bash-5.1$ grep reserved_host_memory_mb /etc/nova/nova.conf
#reserved_host_memory_mb=512
reserved_host_memory_mb=15360
bash-5.1$

From there it can be converted from a derived to a standard paramter like this:

parameter_defaults:
  ComputeHCIParameters:
    NovaReservedHostMemory: 15360

After the parameter has been set as described above it is no longer necessary to deploy the overcloud with the -p option.

The list of available parameters that derived paramters can set is below. The are NovaReservedHostMemory and NovaCPUAllocationRatio are controlled by the HCI workflow while the others are from the NFV worklfows.

  • IsolCpusList

  • KernelArgs

  • NeutronPhysnetNUMANodesMapping

  • NeutronTunnelNUMANodes

  • NovaCPUAllocationRatio

  • NovaComputeCpuDedicatedSet

  • NovaComputeCpuSharedSet

  • NovaReservedHostMemory

  • NovaReservedHostMemory

  • OvsDpdkCoreList

  • OvsDpdkSocketMemory

  • OvsPmdCoreList