I can’t run the jobs in parallel because I’m resource limited on my Selenium hub.
They do *not* depend on each other sequentially.
The use case is that I need a group of jobs to run through to completion in
succession, not parallel, but at the end, if at any point a job had failed, to
fail the build…not ignore failures. There is a guard/rescue for try/finally.
Why not have a try/catch equivalent? That’s basically what I need, I think.
On Jan 10, 2014, at 8:32 AM, nicolas de loof <nicolas.del...@gmail.com> wrote:
> This isn't supported at this time - I don't really get your use-case
> why can't you run those jobs in parallel ? If they actually depend on each
> other sequentially, why not stop the flow when first one fails ?
>
>
> 2014/1/10 silver <pja...@gmail.com>
> Nicolas,
>
> Do you have a recommendation on how I can accomplish the goal at hand?
> Otherwise, I see no other option but to try Marc's groovy script.
>
> Thanks.
>
> On Jan 10, 2014, at 2:51 AM, nicolas de loof <nicolas.del...@gmail.com> wrote:
>
>> I don't recommend such a fully programmatic approach, build-flow is designed
>> as a DSL, admittedly not constrained to just supported keywords (because I
>> didn't know how to do this when I started this plugin) but clearly not
>> supposed to be used to create such a groovy script.
>>
>>
>> 2014/1/10 Marc MacIntyre <marc...@purestorage.com>
>>
>> You are overthinking it :) The trick is to grab the return value from the
>> build() call and check the result of that, then explicitly set the failure
>> state of the buildflow.
>>
>> This is what I'm doing; it's more solution than you need, but it solves your
>> problem.
>>
>> This buildflow takes a map of jobs and the pass criteria, and fires
>> everything off in parallel. If you want to retry on failures, that's
>> supported, and/or you can start several in parallel and pass if some portion
>> of them pass. We use job names as the map key, so if you want to start
>> multiple runs of a particular job with different params, you'll need to
>> modify the script somewhat.
>>
>> def createBuildClosure(String jn, Map args, int retryCount = 0) {
>> // This indirection is needed to force a clone of args, so it's out of
>> scope and gets
>> // re-bound to the closure each time - otherwise jenkins will
>> deduplicate our builds.
>> def ags = args.clone()
>> ags.put("_dedup", java.lang.System.nanoTime())
>> if (retryCount) {
>> return { retry(retryCount) {build(ags, jn)} }
>> } else {
>> return {build(ags, jn)}
>> }
>> }
>>
>> def startParallelRuns(Map buildsToRun) {
>> def m = [:]
>> buildsToRun.each {
>> jobName, params ->
>> def maxFailures = params.get("maxFailures", 0)
>> def retryCount = params.get("retryCount", 0)
>> println "Running "+jobName+" "+params.count+" times (max
>> failures "+maxFailures+")"
>> for (int idx = 0; idx < params.count; idx++) {
>> m.put(jobName+"_"+idx, createBuildClosure(jobName,
>> params.args, retryCount))
>> }
>> }
>>
>> ignore(FAILURE) {
>> join = parallel(m)
>> }
>>
>> results = [:]
>> // process the results by job name
>> buildsToRun.each {
>> jobName, params ->
>> def passcount = 0
>> def maxFailures = params.get("maxFailures", 0)
>> for (int idx = 0; idx < params.count; idx++) {
>> run = join[jobName+"_"+idx]
>> if (run.result == SUCCESS) { passcount += 1}
>> }
>> result = (params.count - passcount) > maxFailures ? FAILURE :
>> SUCCESS
>> println ""+result+": "+jobName+": "+passcount+"/"+params.count+"
>> passed (Max failures: "+maxFailures+")"
>> results[jobName] = result
>> }
>> return results
>> }
>>
>> build_params = params.clone()
>>
>> // Modify your build params here
>> build_params.put('UPSTREAM_JOB', build.project.name)
>>
>> buildsToRun = [
>> job1: [count: 1, maxFailures: 0, args: build_params, retryCount: 2],
>> job2: [count: 1, maxFailures: 0, args: build_params],
>> job3: [count: 1, maxFailures: 0, args: build_params],
>> jobX: [count: 2, maxFailures: 0, args: build_params],
>> jobY: [count: 1, maxFailures: 0, args: build_params],
>> ]
>>
>>
>> results = startParallelRuns(buildsToRun)
>> build.state.result = results.any { job, result -> result == FAILURE} ?
>> FAILURE : SUCCESS
>>
>>
>>
>> On Thu, Jan 9, 2014 at 8:22 PM, silver <pja...@gmail.com> wrote:
>> Sorry for any confusion. The line: ”println(“There were “+FailuresPresent+"
>> test(s) that failed”);" is outside of the if statement resulting in the
>> example output at the end of this message.
>>
>> On Jan 9, 2014, at 9:55 PM, silver <pja...@gmail.com> wrote:
>>
>> > When I run jobs in parallel, the Build Flow fails or passes as I’d expect.
>> > Example:
>> >
>> > parallel (
>> > { build(job1) },
>> > { build(job2) },
>> > { build(job3) },
>> > )
>> >
>> > All of the jobs are started and if they all pass, the Build Flow passes.
>> > If one fails, the Build Flow fails.
>> >
>> > What I’d like to do is to run jobs sequentially, ignoring a failure *for
>> > that moment* but at the end, fail or pass the Build Flow as a whole.
>> > Using “ignore(FAILURE)" doesn’t give me what I want because it will ignore
>> > a failure and pass the Build Flow regardless:
>> >
>> > ignore(FAILURE) {build(job1)}
>> > ignore(FAILURE) {build(job2)}
>> > ignore(FAILURE) {build(job3)}
>> >
>> > If they all fail, the Build Flow still passes because failures are
>> > ignored. But I really need ALL of the jobs to run no matter the outcome
>> > of the other jobs, and the Build Flow to pass/fail, depending on each
>> > outcome.
>> >
>> > Therefore, I have tried something like this (which I thought I got to
>> > actually work at one point but I can’t get it to work again!?! The
>> > closest I can get is explained further down.):
>> >
>> > FailuresPresent = 0;
>> > try {
>> > build(job1)
>> > }catch(e) {
>> > FailuresPresent = FailuresPresent++;
>> > }
>> > try {
>> > build(job2)
>> > }catch(e) {
>> > FailuresPresent = FailuresPresent++;
>> > }
>> > try {
>> > build(job3)
>> > }catch(e) {
>> > FailuresPresent = FailuresPresent++;
>> > }
>> > if ( FailuresPresent>0) {
>> > println(“There were “+FailuresPresent+" test(s) that failed”);
>> > throw new Exception("FAILED!”);
>> > }else {
>> > println "Tests PASSED!";
>> > }
>> >
>> > But the Build Flow will still stop immediately after a failed job (I don’t
>> > see my println at the end). If I use an ignore(FAILURE) wrapper, then the
>> > “catch” is ignored and the Build Flow passes.
>> >
>> > I am not using guard/rescue because I don’t need the FailuresPresent to
>> > increment every time, only when there is a failure (or do I? Guard/Rescue
>> > is like try/finally, not a try/catch.)
>> >
>> > None of my jobs are dependent on another, I just want them all grouped
>> > together and to run sequentially in a single Build Flow if possible.
>> > Running them in parallel maxes out my resources (not Jenkins but my
>> > Selenium hub).
>> >
>> > If I wrap the above jobs in a parallel statement, it seems to gives the
>> > appearance of it finishing to completion (my print statement at the end is
>> > seen) but the Build Flow doesn’t run the other jobs.
>> >
>> > This is the output with the entire try/catch/builds wrapped in a parallel
>> > statement (notice job2 and job3 aren’t run but my println at the end is
>> > seen:
>> >
>> > parallel {
>> > Schedule job job1
>> > Build job1 #34 started
>> > job1 #34 completed : UNSTABLE
>> > }
>> > There were 0 test(s) that failed
>> > Tests PASSED!
>> >
>> > Suggestions? I hope I’m over-thinking this.
>> >
>> > Thanks.
>>
>>
>>
>>
>> --
>> Marc MacIntyre
>>
>> --
>> 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 jenkinsci-users+unsubscr...@googlegroups.com.
>> For more options, visit https://groups.google.com/groups/opt_out.
>>
>>
>> --
>> 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 jenkinsci-users+unsubscr...@googlegroups.com.
>> For more options, visit https://groups.google.com/groups/opt_out.
>
>
>
> --
> 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 jenkinsci-users+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/groups/opt_out.
smime.p7s
Description: S/MIME cryptographic signature
