I have an environment variable injected by Jenkins like:
CUSTOMERS="foo,bar"
Now i need to loop over these values.
Is there any way to access these values AS items in ansible?
Any help including other suggestions how to to solve this is welcome.
You can pass in the environment variable to ansible with --extra-vars, but that's only part of the solution, you need to get the string value into a data format that ansible understands.
One straightforward option is to write a simple Python (or your preferred language) script to convert the environment variable to a JSON list and pass the JSON file to ansible as extra vars with --extra-vars "#customers.json" (JSON file input is available in ansible 1.3), see Ansible Variable documentation.
import sys
import os
import json
DEFAULT_VAR="CUSTOMERS"
def var_to_json(var_name, list_sep = ','):
var_dict = {var_name: os.environ[var_name].split(list_sep)}
return json.dumps(var_dict)
var_name=DEFAULT_VAR
if len(sys.argv) > 1:
var_name = sys.argv[1]
print var_to_json(var_name)
The script above could be generalized further (or customized to your situation). I'll have to leave it to you to hook the pieces together in your build environment.
Alternatively, as this previous answer explains, you can define a custom filter in ansible to process the input value. You could create a filter that converts the variable value to a list, then use the filter in your playbook when referencing the variable (presumably passed in via --extra-vars).
Related
I try to collect data for an Extended Choice Paramater.
I use the code:
def ingo = sh(script: 'mktemp', returnStdout: true)
return ingo
inside the groovy script part, but it seem that this is not allowed or well formated.
The choice is always empty. Anybody has some experience with shell command inside this part of the pipeline?
Sense is, I want to collect data with curl here. But simple shell query is not working.
Please see image
You can't execute Jenkins steps within the Extended Choice Parameter. You have to use pure Groovy within the Parameter. For example, if you want to do an HTTP call you can use a Groovy script like below. Here I'm fetching content from a GitHub file. A Full example can be found here.
def content = new URL ("https://raw.githubusercontent.com/xxx/sample/main/testdir/hosts").getText()
I've run into a specific problem while trying to automate a tagging process in terraform. I've set an environment variable that is essentially a list of all the tags we'd be using for all resources provisioned in the apply. It looks like this...
export TF_VAR_taglist='{JiraEpic = "ETOS-56","AssignedResearcherPri" = "Isaac",AssignedResearcherSec = "Matt"}'
After setting the environment variable I added a variable called "taglist" in the variables.tf file that grabs the aforementioned environment variable. It looks like this...
variable "taglist"{}
Lastly, I have another locals.tf file where i set a common_tags variable. Like so...
locals { common_tags ="${var.taglist}" }
When i terraform apply, the build fails while trying to map the tags properly. This is the error i receive...
Error: Incorrect attribute value type
on kube_master_worker_nodes_ec2.tf line 9, in resource "aws_instance" "master":
9: tags = local.common_tags
|----------------
| local.common_tags is "{JiraEpic = \"ETOS-56\",AssignedResearcherPri = \"Isaac \",AssignedResearcherSec = \"Matt\"}"
Inappropriate value for attribute "tags": map of string required.
I then decided to define the type of the variable as map(string in the variables.tf file like this
variable "taglist"{ type = map(string) }
I had hoped that this would allow terraform to recognize this variable as a map of strings and not just a string literal, but I was wrong, and these are the errors I get when that definition is applied.
Error: Missing attribute separator
on <value for var.taglist> line 1:
(source code not available)
Expected a newline or comma to mark the beginning of the next attribute.
Error: No value for required variable
on variables.tf line 11:
11: variable "taglist"{
The root module input variable "taglist" is not set, and has no default value.
Use a -var or -var-file command line argument to provide a value for this
variable.
I'm really stuck on this, and I feel like I'm close. Can anyone provide some insight into this and how I should go about solving it?
I want to first thank Martin Atkins for giving me the idea of using colons instead of equal signs in my environment variable, because that was the ONLY issue. The variable was not properly represented as a JSON object so terraform was interpreting it as a string.
I changed
export TF_VAR_taglist='{JiraEpic = "ETOS-56","AssignedResearcherPri" = "Isaac",AssignedResearcherSec = "Matt"}'
to this
export TF_VAR_taglist='{"JiraEpic":"ETOS-56","AssignedResearcherPri":"Isaac", "AssignedResearcherSec":"Matt"}'
The build purrs like a kitten on catnip now.
Terraform uses the type constraint of a variable to decide how to interpret a string representation of its value. By default, Terraform will assume the value expects a primitive type such as a string or number, because that's the most typical case for variables set via the command line or environment variables.
Since your tag list is a list you need Terraform to interpret it as a map expression rather than as a string. You can tell Terraform to do that by telling Terraform which type of value you expect:
variable "taglist" {
type = map(string)
}
You can read more about this in the Terraform documentation section Complex-typed Values.
You then need to make sure that the value in the environment variable is a valid object expression in order to avoid a syntax error. If you're setting the environment variable from the shell command line then you need to be mindful of escaping/quoting to ensure that Terraform will see the value with all of the quotes intact, and without any extra metacharacters.
The result is often hard to read clearly, which is why the Terraform documentation recommends using a .tfvars file to set complex-typed variables, instead of the -var command line option or environment variables. However, since you are using automation you might find it easier to generate a .tfvars.json file instead, which uses standard JSON format and is therefore easier to generate using JSON libraries available in most programming languages. Here's the .tfvars.json equivalent of the value you showed in your question:
{
"taglist": {
"JiraEpic": "ETOS-56",
"AssignedResearcherPri": "Isaac",
"AssignedResearcherSec": "Matt"
}
}
Note that subjectively I'd find it pretty confusing to have a variable whose name ends in list when it actually expects a map. A more typical name for this variable would be just tags, though if it's useful to mention its type in order to distinguish it from other variables then I'd suggest tag_map instead, to make it less confusing.
I am trying to create a variable in TFS for my task for timestamp in format yyyymmdd. I know that I create a task specifically for this using bash or powershell. However, I am looking to find some existing variable or some way to create this variable without having to setup a task dedicated for itself.
So far I have tried to use $(Date:yyyymmdd) in my variables but it does not put values in it, it uses variable name as is.
For example, C:\Alpha\beta\$(Date:yyyymmdd) instead of C:\Alpha\beta\20191107
Can anyone help me with this ? Thanks a lot
Actually, it is hence not expanded. We do not have this kind of system or environment variable which get current date time and work every. You may have noticed you could use $(Date), however which is only available in the Build number format section. Others such as $(Rev:r) and $(DateOfYear) are the same, do not work outside the BuildNumberFormat-Settings..
Take a look at the list of all system and environment variables here: Predefined variables
As you have pointed out, you need to use a script in a PowerShell Task to set a variable in your build definition, a sample:
$date=$(Get-Date -Format 'yyyymmdd');
Write-Host "##vso[task.setvariable variable=time]$date"
Then you can use $(time) in your subsequent build tasks.
More details also take a look at this similar question: VSO(TFS) - get current date time as variable
I am trying to pass more than nine parameters from Jenkins to JMeter4.0.
As I was reading, I found out that JMeter does not accept more than 9 parameters. As a workaround, I want to pass all the parameters as a string and split it in JMeter BeanShell.
java -jar -Xms512m -Xmx2048m C:\JMeter4\bin\ApacheJMeter.jar -Jjmeter.save.saveservice.output_format=csv -Jjenkinsparams="%Timetorun%,%Users%" -n -t %JMeterPath%\bin\tests\tests.jmx -l %WORKSPACE%\Results.csv
The tests run on a Windows machine. From this call I have
jenkinsparams = "300,2"
I use a BeanShell PreProcessor like this:
String line = "${__P(jenkinsparams)}";
String[] words = line.split(",");
vars.put("timetorun",words[0]);
vars.put("users",words[1]);
log.info(words[1]);
log.info(users);
I tried few log.info to check the values. For words[1] I have the correct value sent from Jenkins: 2. For the users the value displayed is: void.
I am trying to use it for Number of Threads as: ${__P(users,1)}.
What am I doing wrong? The values clearly arrive from Jenkins but I have a problem passing it to my variable. Thank you
You don't have a script variable named users, so you should either log words[0]:
log.info(words[0]);
Or you can log the value of JMeter variable called users:
log.info(vars.get("users"));
Or you can assign words[0] to variable called users:
String users = words[0];
log.info(users);
Also you are saving it as variable, not property, so you can retrieve it elsewhere in script as
${users}
The syntax __P refers to property, so if you want to use it as property, you need to change how you are saving it:
props.put("users", words[1]);
If you do that, ${__P(users,1)} should work
Now, if you want to use this value as number of threads, then you need to do this:
Have Setup thread group with 1 thread, and your script
In the script you must save users as property, otherwise it won't pass between threads
Next thread group then can use it as number of threads
As long as your command line fits into 8191 characters it should not be a problem to pass as many arguments to JMeter as you want, here is an evidence from Debug Sampler and View Results Tree listener combination
So keep calm and pass as many parameters as needed via -J command line arguments.
Be aware that starting from JMeter version 3.1 users are recommended to use JSR223 Test Elements and Groovy language instead of Beanshell so going forward please consider switching to Groovy.
I want to read a variable from CSV and use that value into another variable.
Example:
I have a variable as:
${url}: wwww.$(value_from_csv}.com
and secondary url ${url}/xyz
In my Jmeter script, ${value_from_csv} is not passed.
What I am missing?
Observed that CSV Dataset Config values are not passed (not available) to any of the Config Elements irrespective of the order of the components (Config Elements) in JMeter Test Plan (checked with User Defined Config & MongoDB Source Config), though passed to Samplers.
so, suggested the OP to define the value in jmeter.properties instead of a CSV file, so we can access user.host in MongoDB Source Config.
Steps:
Add user.host=address in jmeter.properties
Restart Jmeter
Add ${__P(user.host,)} in Server Address List field in MongoDB Source Config
Note: In case of running JMeter script from Jenkins, property will be picked by the script, from jmeter.properites file.
MongoDB Source Config is initialized before any JMeter Variables so the only way to make it dynamic is using JMeter Properties instead.
You can set a JMeter Property in 2 ways:
Define it in user.properties file like:
server.address.1=some.ip.or.hostname.1
server.address.2=some.ip.or.hostname.2
Pass the properties via -J command-line arguments like:
jmeter -Jserver.address.1=some.ip.or.hostname.1 -Jserver.address.2=some.ip.or.hostname.2 ....
See Apache JMeter Properties Customization Guide for more information on using JMeter Properties