https://www.jenkins.io/doc/book/pipeline/scaling-pipeline/ says that durability
is the ability of restarting running jobs even if Jenkins server restarts or
system crashes.It seems not much useful. Jenkins server seems not so easy to
crash.
What are some subtle bugs related to Pipeline durability with nested parallel
steps in the past?I'd like to use need parallel if possible.In some situations,
nested parallel is useful.
For example,you want build artifacts in A,B,C platforms parallelly.Then you run
test artifacts in corresponding A(A1 A2 A3), B(B1 B2 B3), C(C1 C2 C3).If not
using nested parallel, you must wait for A,B,C all finish their building to
start test.If building A cost 10mins, building B cost 40mins, branch A don't
need to wait the remaining 30min.Nested parallel can reduce this kind of
waitting time,branch A just start its own parallel test after branch A finish,
without waitting branch B.In the meantime, if branch A fails, branch B will
continue to run B's test, which is originally not related to branch A's fail.
If you spilt nested parallel into two sequential parallel, you must handle the
fails of different branch, only start test for successful build platforms.Or if
you spilt nested parallel into two Jenkins build jobs,it seems splitting two
related things, which make things harder to understand.
Thanks.
------------------ 原始邮件 ------------------
发件人:
"jenkinsci-users"
<[email protected]>;
发送时间: 2021年11月2日(星期二) 晚上9:52
收件人: "Jenkins Users"<[email protected]>;
主题: Re: thread safety of scripted pipeline parallel and usage of nested
parallel
Scripted Pipeline supports nested parallel steps at the execution level,
although visualizations like Blue Ocean only support one level of
parallelization, and Declarative Pipeline does not natively support nested
parallel, and I do not think nested parallel has much real-world use or
testing. Also, there have been some subtle bugs related to Pipeline durability
with nested parallel steps in the past, so you may want to do some related
testing with your Pipeline before committing to nested parallel if durability
is important to you.
Thanks,
Devin
On Tuesday, November 2, 2021 at 12:42:14 AM UTC-4 abstrakta wrote:
So nested parallel is allowed too ?
---Original---
From: "'Devin Nusbaum' via Jenkins Users"<[email protected]>
Date: Mon, Nov 1, 2021 21:27 PM
To: "jenkinsci-users"<[email protected]>;
Subject: Re: thread safety of scripted pipeline parallel and usage of nested
parallel
The Pipeline execution engine uses a "green thread" model internally - there is
only ever one Java thread executing the Groovy DSL for a given Pipeline (step
execution may use additional Java threads, but this should always be
transparent to users). See the final paragraph under
https://github.com/jenkinsci/workflow-cps-plugin#technical-design for a few
more details. You should not need to worry about Java thread safety across the
various branches in a parallel step as long as you are not doing anything with
Java threads directly in your Pipeline.
On Mon, Nov 1, 2021 at 7:03 AM kuisathaverat <[email protected]> wrote:
Jenkins pipelines have two types of code, the steps and directives provided by
Jenkins, and the Groovy code.
All the Jenkins steps and directives are thread safe. On your Groovy you must
guarantee is thread-safe if it needs to be.
Unstash and archive are Jenkins steps you can use them in parallel tasks and
they must work as expected, I only see one case that archive can cause issues
and is when you archive the same artifact name from different tasks, but is
completely logical that fails or corrupt the file due to you ar doing something
wrong (the same happens with stash in parallel tasks)
The last thought, try to minimize all the groovy code you add to your
pipelines, all Groovy code run in the Jenkins controller, try to delegate on
scripts, commands, tools, ... as much as possible.
El dom, 31 oct 2021 a las 18:55, 'abstrakta' via Jenkins Users
(<[email protected]>) escribió:
sorry, pipeline code should be correct like this:
pipeline {
agent any
stages {
stage('Parallel Build') {
steps {
script {
def
builds = [:]
def
tests = [:]
stash name: 'src', include: 'src/**'
def
build_action = { platform, tests ->
//build in some slave node.
node ("${platform}") {
//unstash need thread mutex lock?
unstash name: 'src'
sh "make ${platform}"
//archive need thread mutex lock?
archive name: "./${platform}"
}
//nested parallel, outside of node block
parallel tests
}
def
test_action = { platform ->
//test in some slave node.
node ("${platform}") {
sh "make ${platform}-test"
}
}
tests["A1"] = { test_action("A1") }
tests["A2"] = { test_action("A2") }
tests["A3"] = { test_action("A3") }
builds["A"] = { build_action("A",tests) }
tests = [:]
tests["B1"] = { test_action("B1") }
tests["B2"] = { test_action("B2") }
builds["B"] = { build_action("B",tests) }
tests = [:]
tests["C1"] = { test_action("C1") }
tests["C2"] = { test_action("C2") }
builds["C"] = { build_action("B",tests) }
parallel builds
}
}
}
}
}
---原始邮件---
发件人: "'abstrakta' via Jenkins Users"<[email protected]>
发送时间: 2021年10月31日(周日) 晚上10:48
收件人: "jenkinsci-users"<[email protected]>;
主题: 回复： thread safety of scripted pipeline parallel and usage of nested parallel
OK,I know what you means. But I test these pipeline that data is not
synchronized many times( about more than 700 times,I use periodical build to
test it automatically). Shared data that is not synchronized is processed
correctly every time. Do you really encounter the situation that shared data is
corrupt in parallel task?
In some scenarios, I want to parallel build one more artifacts that used in
different platforms from the same source code. And then I parallel test all
artifacts in their corresponding platforms(One artifact might be used in one
more platforms).
build platform: A B C
test platform: A(A1 A2 A3), B(B1 B2), C(C1 C2)
-> A1
-> A -> A2
|
-> A3
/
source code --> B -> B1
\ -> B2
|
-> C -> C1
-> C2
-> C3
Pipeline code like this:
pipeline {
agent any
stages {
stage('Parallel Build') {
steps {
script {
def
builds = [:]
def
tests = [:]
stash name: 'src', include: 'src/**'
def
build_action = { platform, tests ->
//build in some slave node.
node ("${platform}") {
//unstash need thread mutex lock?
unstash name: 'src'
sh "make ${platform}"
//archive need thread mutex lock?
archive name: "./${platform}"
}
//nested parallel, outside of node block
parallel tests
}
def
test_action = { platform ->
//test in some slave node.
node ("${platform}") {
sh "make ${platform}-test"
}
}
tests["A1"] = test_action("A1")
tests["A2"] = test_action("A2")
tests["A3"] = test_action("A3")
builds["A"] = build_action("A",tests)
tests = [:]
tests["B1"] = test_action("B1")
tests["B2"] = test_action("B2")
builds["B"] = build_action("B",tests)
tests = [:]
tests["C1"] = test_action("C1")
tests["C2"] = test_action("C2")
builds["C"] = build_action("B",tests)
parallel builds
}
}
}
}
}
I have been running this pipeline for some time. It works well. But I worry
about the thread safety of unstash(and archive，or other commands). I want to
find some evidence that Jenkins garantee these parallel thread safety and
nested parallel is ok. Otherwise I must split these platforms into different
build jobs(which I think it's more difficult to manage projects.) Because
Jenkins must garantee parallel build jobs more safe? I just don't know the
underlying works that Jenkins do about parallel.I'm just curious about these
thread safety.If Jenkins take care of these synchronization,that is the best.
If such, I don't need to care about these thread things. If not,I want to find
correct ways to do parallel tasks.
------------------ 原始邮件 ------------------
发件人:
"jenkinsci-users"
<[email protected]>;
发送时间: 2021年10月31日(星期天) 晚上7:19
收件人: "Jenkins Users"<[email protected]>;
主题: Re: thread safety of scripted pipeline parallel and usage of nested
parallel
The following example should work (I did not test it), I my case I have used
maps like the “result” variable, that it is a simple map not synchronized and
stores data from all task but it is not read from the different tasks. The
other two cases “data” and “mapSync” uses concurrent classes, they are thread
safe and synchronized so you can share data across tasks, I dunno is they are
in the allow list for pipeline, if not you have to approve it use in pipelines
in the Jenkins config. Finally, the last part of the pipeline uses nested
parallel task, from my experience is not a good idea, the parallel explosion of
task is a little incontrolable and there are other solution like launch a job
from those task and inside that job launch parallel task, in this way you only
have 1 parallel level that is easy to control and understand when something
goes wrong.
import groovy.transform.Field
import java.util.concurrent.atomic.AtomicInteger
import java.util.concurrent.ConcurrentHashMap
@Field def results = [:]
@Field AtomicInteger data = new AtomicInteger()
@Filed ConcurrentHashMap mapSync = new ConcurrentHashMap<String,Integer>()
pipeline{
agent any
stages
{
stage(‘Parallel BuiLd’) {
steps {
script {
def i = 0
def
builds = [:]
mapSync[“odd”] = 0
mapSync[“even”] = 0
stash
name: ‘src’, include: ‘src/**’
//generate 1000 parallel block
for (i =
0; i<1000; i++) {
// make the Map of Closure
builds[“$i”] = {
results[“$i”] = 1
data++
if(i%2==0){
mapSync[“odd”] = mapSync[“odd”]++
} else {
mapSync[“even”] = mapSync[“even”]++
}
}
}
parallel
builds
println
results.toString()
println
data
println
mapSync
}
}
}
}
}
El sábado, 30 de octubre de 2021 a las 18:44:57 UTC+2, abstrakta escribió:
Thanks for your reply.
So parallel directive is like spawning some Java threads?Do you have any
pipeline code can demo this thread safety issue and how to fix it using Java
types that are thread safe ?
I guess that some directive "stash unstash archive" should be thread safety
internally.Because I find some articles use parallel unstash in different slave
node without thread protection.Is my guess correct?I can't find any other
articles that discuss this parallel thread safety issue.
---Original---
From: "Ivan Fernandez Calvo"<[email protected]>
Date: Sun, Oct 31, 2021 00:03 AM
To: "Jenkins Users"<[email protected]>;
Subject: Re: thread safety of scripted pipeline parallel and usage of nested
parallel
No, if you plan to use shared variables across parallel steps you should use
Java types that are thread safe, if not you will have random/weird results. I
have several pipelines that uses a map to store results in parallel steps
El sábado, 30 de octubre de 2021 a las 14:18:29 UTC+2, abstrakta escribió:
Hi, Jenkins friends.I wish that I'm in the right place that post these Jenkins
usage question.
I find that the Scriped Pipeline parallel works like threads.Unlike the
Declarative Pipeline parallel,the Scriped Pipeline parallel just use one
executor.Closure in parallel parameters works like a thread.
My question is:
1.Does Jenkins garantee the data thread safety of parallel closure internally?
2.Does I need to care about the thread safety of the commands that executes in
scriped parallel closure?
3.Is there any limit of usage in the commands that executes in parallel?Can I
use nested scripted parallel? Why the documentation of Declarative Pipeline
parallel in Pipeline Syntax reference says that "Note that a stage must have
one and only one of steps, stages, parallel, or matrix. It is not possible to
nest a parallel or matrix block within a stage directive if that stage
directive is nested within a parallel or matrix block itself."
I test some nested Pipeline code that might cause thread race condition many
times.Jekins always give the right answer that shared data is modified
correctly.Is this thread safety garanteed in the design of Jenkins parallel
directive?
Pipeline code like this:
pipeline{
agent any
stages
{
stage('Parallel BuiLd') {
steps {
script {
def
i = 0
def
data = 0
def
builds = [:]
stash name: 'src', include: 'src/**'
//generate 1000 parallel block
for
(i = 0; i<1000; i++) {
// make the Map of Closure
builds["$i"] = {
//modify shared data, need thread mutex lock?
data++
//unstash or other command, need thread mutex lock?
unstah name: 'src'
def tests = [:]
// ... generate tests Map
// Can I use nested parallel?
parallel tests
}
}
parallel builds
println data //It does always print 1000
}
}
}
}
}
The variable data is always modified to 1000. So Jenkins garantee the thread
safety of parallel?
--
You received this message because you are subscribed to the Google Groups
"Jenkins Users" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/jenkinsci-users/0712b8c5-42dc-439f-a017-2a5ca45ad1e9n%40googlegroups.com.
--
You received this message because you are subscribed to the Google Groups
"Jenkins Users" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/jenkinsci-users/fed01b0e-a04c-448b-9f32-3d2b3ffd6c40n%40googlegroups.com.
--
You received this message because you are subscribed to the Google Groups
"Jenkins Users" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/jenkinsci-users/tencent_579508DE2429C25B92D7BFFF6AED09115C08%40qq.com.
--
You received this message because you are subscribed to a topic in the Google
Groups "Jenkins Users" group.
To unsubscribe from this topic, visit
https://groups.google.com/d/topic/jenkinsci-users/9fzFnwLkfxU/unsubscribe.
To unsubscribe from this group and all its topics, send an email to
[email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/jenkinsci-users/tencent_E81808632457C23792EEBFC46A5A093E5C08%40qq.com.
--
Un SaludoIván Fernández Calvo
https://www.linkedin.com/in/iv%C3%A1n-fern%C3%A1ndez-calvo-21425033
--
You received this message because you are subscribed to the Google Groups
"Jenkins Users" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/jenkinsci-users/CAKo5Qrq%2BP9qYn%3Dm9Ssc4aSCAkedr_7D1jRTSuKYPQNxJ_3xrmw%40mail.gmail.com.
--
You received this message because you are subscribed to the Google Groups
"Jenkins Users" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/jenkinsci-users/CAF73ShC883ws%3DYvMVcBeeR%3Dyub8iyjA0%3DidL_pzDocki0Rp5dg%40mail.gmail.com.
--
You received this message because you are subscribed to the Google Groups
"Jenkins Users" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/jenkinsci-users/e6017246-339c-4bd0-8197-bf5abb05bb8bn%40googlegroups.com.
--
You received this message because you are subscribed to the Google Groups
"Jenkins Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/jenkinsci-users/tencent_3947923BE10CC99A5F2035B68479AE6D0407%40qq.com.
