4. Template generation¶
In order to create an application from a Robin bundle one needs to provide an appropriate application template JSON. Robin allows one to natively generate a template for a specified bundle to ease this process. Often run-time configurations/parameters need to be specified during application deployment such that they override the defaults that present in the bundle manifest file. As a result, one can pass in an optional input YAML wherein these modifications are specified so as to create template with these customized values in place.
Topics covered in this section include:
|
Generate a application template from a bundle |
4.1. User Input YAML¶
A definitive input YAML is shown below with all the possible values that can be specified. Comments detailing the purpose of each variable are provided in-line.
# This is a sample YAML file which can be used as an input to generate
# a ROBIN template from a bundle to create ROBIN applications.
# Use only the keys that are required for your application.
# Name of the app
name: "appname"
# Resource pool
rpool: "default"
# VNC required ?
vnc: false
# Set restart count for Robin AP to try re-deploying VNodes
# for failed applications
restart_num: 5
# Array of IPPools
ippools: ["robin-default"]
# Appvars
appvars:
json:
key1: "value1"
key2: "value"
# Resource groups
resgrps:
- name: scan1
count: 1
attributes: {}
- name: scan2
count: 1
attributes: {}
# Set snapshot schedule for app
snapshot_schedule:
cron: 10 * * * *
retain: 2
# App is numa aware
numa_aware: null
# Docker Registry Name (Can be overriden at the role level)
registry: "Internal"
# Default replication factor (Can be overriden at the role level)
replication: 1
# Default faultdomain (Can be overriden at the role level)
faultdomain: null
# NO rootfs for the entire app (Can be overriden at the role level)
norootfs: false
# Rootfs size for the entire app (Can be overriden at the role level)
rootfs_size: null
# Media for the entire app (Can be overriden at the role level)
media: HDD
# App snapshot space percent for the entire app (Can be overriden at the role level)
snap_reserved: 20
# Add global option to skip creation of K8S headless service as a part of Robin App creation
skip_headless_service: True
# Set the network policy scope for an application. Valid values include: "allow-all", "namespace", "tenant", "user", "app", "deny-all"
# If not specified the default value is "allow-all"
network_policy_scope: "namespace"
# ROLES section. An array of roles.
# Please make sure the role names match exactly the role names
# present in the application bundle being used
roles:
- name: role1 # Replace this with an actual role in the bundle
# Number of vnodes in the Role
multinode_value: 2
# Number of Vnode restart attempts to be done by Robin AP
# This role level setting overrides the global restart_num setting listed above
restart_num: 2
# No root fs for this role
norootfs: true
# root fs size for this role
rootfs_size: "10G"
# If role has candisable in bundle and user wants to disable role1
disabled: true
# Add role level option to skip creation of K8S headless service as a part of Robin App creation
# This option will overide the global level skip_headless_service option
skip_headless_service: True
# dual_stack option can be used for OVS dual stack clusters to have a 1:1 mapping between
# pod hostnames and IP addresses
dual_stack: True
# shareProcessNamespace option can be used to enable process namespace sharing where processes in a container
# are visible to all other containers in that pod.
shareProcessNamespace: True
# skip_global_ippool option allows users to skip adding global IP pool to role vnodes network section
# during application creation. By default, Robin will add the global ippool to all role's vnode network section
skip_global_ippool: True
# Set the network policy scope for vnodes created as part of this role.
# Valid values include: "allow-all", "namespace", "tenant", "user", "app", "deny-all"
# This role level option will override the value for the global setting for this option
network_policy_scope: namespace
# Role specific ippools with customizable parameters
# These are seperate from the global IP pool which is defined as the first ippool at the app level
# name (optional) parameter per ip pool will be used to map Robin IP pool names
# to NET names defined in the bundle manifest YAML files
# primary (optional) parameter per ip pool will be used to support Robin feature to configure
# multiple IPs on a single pod interface. This feature is only supported with OVS and SR-IOV IP pools
# trust (optional) parameter per SR-IOV ip pool will be used to have trust enabled for SR-IOV VFs Default: off
# spoofchk (optional) parameter per SR-IOV ip pool will be used to have sppofchk enabled/disabled for SR-IOV VFs. Default: on
# bond_mode (optional) parameter per SR-IOV ip pool will be used to set bonding mode. Used with bond CNI. Default: active-backup
# mtu (optional) parameter will be used to set custom MTU configuration per IP pool
# staic_macs (optional) per ip pool will be used to set static MAC addresses for each Pod/Vnode
# static_ips (optional) per ip pool will be used to set static IP addresses for each Pod/Vnode
# tags (optional) per SR-IOV ip pool will be used to select NICs based on key/value pair passed
ippools:
- ippool: private
static_ips: ["192.168.1.101", "192.168.1.102"]
static_macs: ["02:0b:0c:11:22:33", "02:0b:0c:11:32:44"]
name: fronthaul
trust: on
bond_mode: active-backup
spoofchk: off
mtu: 1500
- ippool: pool-2
name: midhaul
trust: on
bond_mode: active-backup
spoofchk: off
mtu: 1500
tags:
name: sriov0
- ippool: pool-3
name: fcaps
- ippool: pool-4
static_ips: ["10.168.1.101"]
- ippool: pool-5
static_ips: ["20.168.1.101", "20.168.1.102"]
- ippool: test-primary-pool
- ippool: test-secondary-pool
primary: test-primary-pool
# Tolerations to be added to VNodes within a role for an application
# to tolerate node taints.
# This can be used to add new tolerations to a bundle manifest role
# apart from any tolerations already specified in the bundle.
tolerations:
- effect: NoSchedule
key: node-role.kubernetes.io/master
- effect: NoSchedule
key: node-role.kubernetes.io/agent
# Labels to be added to VNodes within a role for an application.
# This can be used to add new labels to a bundle manifest role
# apart from labels already specified in the bundle
labels:
labelkey1: labelvalue1
labelkey2: labelvalue2
# Compute inputs for the role.
# Value for terminationGracePeriodSeconds to be used in K8S pod spec
grace_period_seconds: 60
containers:
# This is the main container without name
- image: "imgname:imgversion"
imagePullPolicy: "IfNotPresent"
cpu:
isol: true # Used for isol-shared/isol-dedicated workloads
nonisol: true # Used for non-isol workloads
reserve: true # Used for isol-dedicated
min: 2.5 # Used for minimum number of CPU cores
max: 2.5 # Used for maximum number of CPU cores
memory: "8G"
hugepages-1Gi: "0"
hugepages-2Mi: "0"
gpu:
count: 2
type: "nvidia-a100"
# Env customization for the main container.
env:
key5: value5
key6: value6
# Specify container <-> host ports mapping (only for main app container pod spec)
ports:
- containerPort: 80
hostPort: 8099
protocol: TCP
- name: side1
image: "imgname:imgversion"
imagePullPolicy: "IfNotPresent"
cpu: "8"
memory: "8G"
hugepages-1Gi: "0"
hugepages-2Mi: "0"
# Env customization for the sidecar container.
# Users should be able to update existing ENV variables
# as well as add new ENV variables for sidecar containers
env:
key1: value1
key2: value2
- name: init1
image: "imgname:imgversion"
imagePullPolicy: "IfNotPresent"
cpu: "8"
memory: "8G"
hugepages-1Gi: "0"
hugepages-2Mi: "0"
# Env customization for the init container.
# Users should be able to update existing ENV variables
# as well as add new ENV variables for init containers
env:
key3: value3
key4: value4
# Hostnames for the vnodes in this role inputs for the role.
hostnames: ["h1", "h2"]
# Registry to use for this role.
registry: null
# Role and Vnode level service definitions
# Only existing service definitions in Robin bundles will be updated
# as part of services section here. No new services will be added to the bundle
services:
- type: NodePort
labels:
l1: v1
l2: v2
annotations:
a1: av1
a2: av2
scope: pod
ports:
- port: 80
protocol: TCP
name: web
node_port: 30000
- type: ClusterIP
labels:
l1: v1
l2: v2
annotations:
a1: av1
a2: av2
ports:
- port: 80
protocol: TCP
name: web
# Storage customization per volume type. One entry per voltype
storage:
data:
count: 3
blocksize: 4096
compression: "disabled"
fstype: "ext4"
size: "10G"
snap_reserved: 20
encryption: "chacha20"
replication: 3
faultdomain: "host"
fstype: "ext4"
media: "HDD"
qos_enabled: False,
qos_priority: 1,
qos_rd_max_iops: 2000000,
qos_rd_min_iops: 0,
qos_rd_min_window: 500,
qos_rd_weight: 1,
qos_wr_max_iops: 2000000,
qos_wr_min_iops: 0,
qos_wr_min_window: 500,
qos_wr_weight: 1
log:
count: 3
size: 2G
# Placement constraints for this role. One of the following 4 can be given
placeon_different_nodes_on_same_rack: false
placeon_different_nodes_on_different_rack: false
round_robin_across_different_racks: false
round_robin_across_different_datacenters: false
# Antiaffinity for this role. One of the following 3 can be enabled
anti_affinity_on_host: false
anti_affinity_on_rack: false
anti_affinity_on_datacenter: false
# Compute and storage affinity for this role. One of the 3 can be enabled
compute_storage_on_same_node: false
compute_storage_on_same_rack: false
compute_storage_on_same_datacenter: false
# Place the vnodes of this role matching host tags
placeon_host_match_tags:
worker: "worker2"
floor: "first"
# Colocate this role with other roles in the application. One of the 3 can be enabled
colocate_with_roles_on_host: ["role2", "role3"]
colocate_with_roles_on_rack: ["role4", "role5"]
colocate_with_roles_on_datacenter: ["role6"]
# Do colocate this role with other roles in the application. One of the 3 can be enabled
donot_colocate_with_roles_on_host: ["role4", "role5"]
donot_colocate_with_roles_on_rack: ["role6"]
donot_colocate_with_roles_on_datacenter: []
# Role 2
- name: role2 # Replace this with an actual role in the bundle
multinode_value: 2
# Role 3
- name: role3 # Replace this with an actual role in the bundle
multinode_value: 2
# Role 4
- name: role4 # Replace this with an actual role in the bundle
multinode_value: 2
Note
It is not necessary to include all roles in the bundle as a part of the input file; one can customize values for particular roles and use default bundle values for the others.
4.2. Generating a Template¶
In order to generate an application template for a specific bundle with run time configurations, issue the following command:
# robin template generate <bundleid>
--input <input_yaml>
--details
--table
|
Id of Application Bundle to generate a template for |
|
Input file containing custom run time values |
|
Display the resulting template as it would be shown in the Robin UI |
|
Display the resulting template in a tabular format |
Example 1 (Custom values for multiple roles):
# cat user-input-demo.yaml
appname: "elk1"
roles:
- name: data_node
multinode_value: 5
containers:
- cpu: 8
memory: 8G
# robin template generate 3 --input user-input-demo.yaml --table
Role | Engine | Image | Vnodes | Cpu | Mem | Network | Storage(Vols/Size)
----------------------+--------+-----------------------------------------------------+--------+--------+----------+---------------+--------------------
data_node | docker | docker.elastic.co/elasticsearch/elasticsearch:6.6.2 | 5 | 8 (40) | 8G (40G) | robin-default | 1/200G (5/1000G)
dedicated_master_node | docker | docker.elastic.co/elasticsearch/elasticsearch:6.6.2 | 1 | 1 (1) | 1G (1G) | robin-default | 1/200G (1/200G)
kibana | docker | docker.elastic.co/kibana/kibana:6.6.2 | 1 | 1 (1) | 1G (1G) | robin-default | -
logstash | docker | docker.elastic.co/logstash/logstash:6.6.2 | 1 | 1 (1) | 1G (1G) | robin-default | 1/10G (1/10G)
master_eligible_node | docker | docker.elastic.co/elasticsearch/elasticsearch:6.6.2 | 1 | 1 (1) | 1G (1G) | robin-default | 1/200G (1/200G)
| | | | | | |
Total | | | 9 | 44 | 44G | | 8/1T
Example 2 (Custom values for networking attributes):
# cat input-static-ips-pools.yaml
appname: "ovs-staticips-app"
media: HDD
ippools: ["ovs-1"]
roles:
- name: server
multinode_value: 2
ippools:
- ippool: ovs-1
- ippool: ovs-2
static_ips: ["10.168.10.122", "10.168.10.123"]
- ippool: ovs-3
static_ips: ["20.168.20.122", "20.168.20.123"]
static_macs: ["02:0b:0c:11:22:33", "02:0b:0c:11:32:44"]
# robin template generate 1 --input-static-ips-pools.yaml --table
Role | Engine | Image | Vnodes | Cpu | Mem | Network | Storage(Vols/Size)
-------+--------+-------------------+--------+-------+-------------+-------------------+--------------------
server | docker | robinsys/centos:7 | 2 | 1 (2) | 200M (400M) | ovs-1,ovs-2,ovs-3 | 2/2G (4/4G)
| | | | | | |
Total | | | 2 | 2 | 400M | | 4/4G
Note
Users can also create applications by directly specifying the input YAML file in place of the template JSON in the robin app create from-bundle command.