Writing recipes
Refer to these guidelines when creating your recipes.
When using recipes, consider the following guidelines:
- Running bash and python scripts as recipes is supported. We recommend
using scripts with Shebang character sequence, for
example:
#!/bin/sh #!/bin/bash #!/usr/bin/sh #!/usr/bin/bash #!/usr/bin/env sh #!/usr/bin/env bash #!/bin/sh -x #!/usr/bin/python #!/usr/bin/env python
- The scripts are executed as root. The recipe output is written to
/var/log/recipes
on each node on which it was executed. -
Supported parameters can be specified as variables by using mustache kind of templating with "{{{ }}}" syntax. Once specified in a recipe, these variables are dynamically replaced when the recipe is executed, generating the actual values that you provided as part of cluster creation process. For the list of parameters, refer to Recipe and cluster template parameters. For an example, see Example: Recipe using parameters.
For example, if your cluster includes an external LDAP and your recipe includes
{{{ldap.connectionURL}}}
, as demonstrated in the following example#!/bin/bash -e main() { ping {{{ ldap.connectionURL }}} } [[ "$0" == "$BASH_SOURCE" ]] && main "$@"
then, when this recipe runs, the
{{{ldap.connectionURL}}}
is replaced with the actual connection URL specified as part of cluster creation process, as demonstrated in the following example:#!/bin/bash -e main() { ping 192.168.59.103 } [[ "$0" == "$BASH_SOURCE" ]] && main "$@"
- Recipe logs can be found at
/var/log/recipes/${RECIPE_TYPE}/${RECIPE_NAME}.log
- The scripts are executed on all nodes of the host groups that you select (such as “master”, “worker”, “compute”).
- In order to be executed, your script must be in a network location which is accessible from the Cloudera Management Console and the virtual network in which your cluster is located.
- Make sure to follow Linux best practices when creating your scripts. For example, don’t forget to script “Yes” auto-answers where needed.
- Do not execute
yum update –y
as it may update other components on the instances (such as salt) – which can create unintended or unstable behavior.
Example Python script
#!/usr/bin/python
print("An example of a python script")
import sys
print(sys.version_info)
Example bash script for yum proxy settings
#!/bin/bash
cat >> /etc/yum.conf
<<ENDOF
proxy=http://10.0.0.133:3128
ENDOF
Example recipe including variables
Original recipe:
#!/bin/bash -e
function setupAtlasServer() {
curl -iv -u {{{ general.userName }}}:{{{ general.password }}} -H "X-Requested-By: ambari" -X POST -d '{"RequestInfo":{"command":"RESTART","context":"Restart all components required ATLAS","operation_level":{"level":"SERVICE","cluster_name":"{{{ general.clusterName }}}","service_name":"ATLAS"}},"Requests/resource_filters":[{"hosts_predicate":"HostRoles/stale_configs=false&HostRoles/cluster_name={{{ general.clusterName }}}"}]}' http://$(hostname -f):8080/api/v1/clusters/{{{ general.clusterName }}}/requests
}
main() {
setupAtlasServer
}
[[ "$0" == "$BASH_SOURCE" ]] && main "$@"
Generated recipe (to illustrate how the variables from the original recipe were replaced during cluster creation):
#!/bin/bash -e
function setupAtlasServer() {
curl -iv -u admin:admin123 -H "X-Requested-By: ambari" -X POST -d '{"RequestInfo":{"command":"RESTART","context":"Restart all components required ATLAS","operation_level":{"level":"SERVICE","cluster_name":"super-cluster","service_name":"ATLAS"}},"Requests/resource_filters":[{"hosts_predicate":"HostRoles/stale_configs=false&HostRoles/cluster_name=super-cluster"}]}' http://$(hostname -f):8080/api/v1/clusters/super-cluster/requests
}
main() {
setupAtlasServer
}
[[ "$0" == "$BASH_SOURCE" ]] && main "$@"