[JENKINS-27948] Script timeout during start of configure a jenkins job

[timja]

I updated jenkins from version 1.596 to version 1.609.
No I get a script timeout if I try to configure some of the larger jobs.
I can configure small jobs without any problems.
I increase the firefox script timeout from 10 seconds to 100 seconds.
Now I can configure the jenkins jobs, but it takes a very long time to start editing.
With version 1.596 it was possible to start configuring of the same larger jobs in the predefined timout of 10 seconds

---

Originally reported by rw250155, imported from: Script timeout during start of configure a jenkins job

• status: Resolved
• priority: Minor
• resolution: Fixed
• resolved: 2016-07-11T14:05:54+00:00
• imported: 2022/01/10

[timja]

danielbeck:

Please provide a list of installed plugins and their versions, as well as a sample config.xml of such a 'large job'.

[timja]

rw250155:

The list of the installed plugins (copied from display of the system properties):

ant 1.2 true true
antisamy-markup-formatter 1.3 true true
any-buildstep 0.1 true false
artifactdeployer 0.33 true false
audit-trail 2.1 true false
backup 1.6.1 true false
batch-task 1.17 true false
build-blocker-plugin 1.6 true false
build-environment 1.4 true false
build-pipeline-plugin 1.4.7 true false
build-user-vars-plugin 1.4 true false
clone-workspace-scm 0.6 true false
conditional-buildstep 1.3.3 true false
config-file-provider 2.7.5 true false
copy-to-slave 1.4.4 true false
copyartifact 1.35 true false
credentials 1.22 true true
custom-tools-plugin 0.4.4 true false
CustomHistory 1.6 true false
cvs 2.12 true true
dashboard-view 2.9.4 true false
delivery-pipeline-plugin 0.8.11 true false
dependencyanalyzer 0.7 true false
deploy 1.10 true false
disk-usage 0.25 true false
downstream-buildview 1.9 true false
downstream-ext 1.8 true false
dynamic-axis 1.0.3 true false
dynamic_extended_choice_parameter 1.0.1 true false
dynamicparameter 0.2.0 true false
email-ext 2.39.3 true false
envfile 1.2 true false
envinject 1.91.2 true false
extended-choice-parameter 0.35 true false
extensible-choice-parameter 1.2.2 true false
external-monitor-job 1.4 true true
flexible-publish 0.15.1 true false
git 2.3.5 true false
git-client 1.16.1 true false
git-server 1.6 true false
groovy 1.24 true false
groovy-label-assignment 1.1.0 true false
groovyaxis 0.3 true false
ivy 1.24 true false
javadoc 1.3 true true
job-log-logger-plugin 1.0 true false
job-node-stalker 1.0.3 true false
jobConfigHistory 2.10 true false
jquery 1.11.2-0 true false
jquery-ui 1.0.2 true false
junit 1.5 true true
labeled-test-groups-publisher 1.2.7 true false
ldap 1.11 true true
m2-repo-reaper 1.0 true false
mailcommander 1.0.0 true false
mailer 1.15 true true
managed-scripts 1.1.1 true false
mantis 0.26 true false
mapdb-api 1.0.6.0 true false
matrix-auth 1.2 true true
matrix-combinations-parameter 1.0.7 true false
matrix-project 1.4.1 true false
Matrix-sorter-plugin 1.1 true false
maven-plugin 2.9 true true
metadata 1.1.0b true false
metrics 3.0.9 true false
monitoring 1.55.0 true false
nodelabelparameter 1.5.1 true false
pam-auth 1.2 true true
parameterized-trigger 2.26 true false
periodicbackup 1.3 true false
postbuild-task 1.8 true false
postbuildscript 0.17 true false
PrioritySorter 2.11 true false
role-strategy 2.2.0 true false
run-condition 1.0 true false
scm-api 0.2 true false
script-security 1.13 true false
scriptler 2.7 true false
selection-tasks-plugin 1.0 true false
shared-objects 0.44 true false
show-build-parameters 1.0 true false
skip-certificate-check 1.0 true false
sonar 2.2 true false
ssh-credentials 1.11 true true
ssh-slaves 1.9 true true
statusmonitor 1.3 true false
strawboss 1.3 true false
subversion 2.5 true true
support-core 2.20 true false
svn-tag 1.17 true false
testlink 3.10 true false
throttle-concurrents 1.8.4 true false
token-macro 1.10 true false
translation 1.12 true true
uno-choice 0.24 true false
update-sites-manager 1.0.1 true false
validating-string-parameter 2.3 true false
vsphere-cloud 2.4 true false
weblogic-deployer-plugin 3.2 true false
websphere-deployer 1.3-SNAPSHOT (private-08/05/2014 09:34-nothhard) true false
windows-slaves 1.0 true false
ws-cleanup 0.25 true false
xunit 1.94 true false

[timja]

rw250155:

The attached zip file contains the config.xml file from a job called DOXiS4_CSB-Test-All-Comb".
Opening this job for configuration exceeds the limit of 100 seconds configured in my firefox-browser.
During this time one core of the quad-core-cpu is busy for the whole time.

[timja]

rw250155:

I increased the timeout value for scripts to run to 200 seconds.
The start to configure the job I added in the attachments took 180 seconds (yes 3 minutes) on my machine.

[timja]

rw250155:

If I should present the data in a special way or format, please feel free to contact me.

[timja]

danielbeck:

Haven't had time to investigate, but you could do some of that yourself:
While the form is submitted and the page is loading, repeatedly load the /threadDump URL and look for a thread named Handling POST /job/JOBNAME/configSubmit or something similar. Look where it's stuck processing for most of the time.

[timja]

rw250155:

Excuse my late answer.
I am glad that I can help.
I started configuring the job I added in this call as example.
The browser was about 3 minutes in the state loading data.
At this time a call to the threadDump URL does not show a thread "Handling POST" or "configSubmit".
The only Thread with Hanling was the threadDump itself.

It look like jenkins generates Javascript, which runs all the time in the browser.
It seems it runs a lot longer, if there a more objects or lines to display.

[timja]

rw250155:

Skript: http://sst-bld-master.sst.loca…/0da8a904/scripts/prototype.js:2188

It seems the script is calculating the size of the objects to display.

[timja]

rw250155:

Depending on the time, I configure as timeout value for the time a script may run, there are different javascripts active at the time the browser stops the script.

[timja]

raspy:

I observe the same issue also in a different scenario: being in a job with a large list of completed builds, clicking on 'more...' at the bottom of the build list results in the same annoying popups with timeouts in prototype.js. Firefox developer tools console logs:

Error: Script terminated by timeout at:
Element.Methods.removeClassName@https://server:8080/static/63938483/scripts/prototype.js:2371:9
getElementOverflowParams@https://server:8080/static/63938483/scripts/hudson-behavior.js:1952:5
checkRowCellOverflows@https://server:8080/static/63938483/scripts/hudson-behavior.js:1693:37
checkAllRowCellOverflows@https://server:8080/static/63938483/scripts/hudson-behavior.js:1883:13
updateBuilds/<.onSuccess@https://server:8080/static/63938483/scripts/hudson-behavior.js:1927:21
Ajax.Request<.respondToReadyState@https://server:8080/static/63938483/scripts/prototype.js:1655:10
Ajax.Request<.onStateChange@https://server:8080/static/63938483/scripts/prototype.js:1600:7
bind/<@https://server:8080/static/63938483/scripts/prototype.js:414:14

[timja]

danielbeck:

same issue also in a different scenario

In other words, a completely different issue: JENKINS-26445

[timja]

raspy:

Thank you, I could not find it However, I actually get both issues, JENKINS-26445 and JENKINS-27948, so I would suggest they might be related.

[timja]

brianeray:

Our upgrade to 1.609.1 seems to be exhibiting the same JS timeouts on an admittedly too-big job config, regardless of the number of builds. The timeouts usually happen in codemirror.js. New to JIRA and am wondering if adding a copy of the config scrubbed of proprietary information will help. I can gather other information about our install as well: plugins, client and server attributes, etc.

Luckily the workaround of setting JS timeouts in Firefox is working. (32 bit Windows v38.0.5, all three dom.max_*_script_run_time params set to 120.) The page is coming up somewhere in the neighborhood of 100 seconds.

If it would help to attach our bits of evidence, let me know.

[timja]

dilipm79:

I can configure small jobs without any problems.
I increase the firefox script timeout from 10 seconds to 100 seconds.
Now I can configure the jenkins jobs, but it takes a very long time to start editing.

Same issue with LTS 1.609.1 & LTS 1.625.1 too.

[timja]

scoheb:

We were able to workaround this by doing the following:

Navigate to:

• Manage Jenkins
• Configure Global Security
• In the Markup Formatter section, ensure:
  • Safe HTML is selected
  • Click on Disable syntax highlighting

With this change, we went from a config page load that would timeout in Firefox to one that loads in under 20s.

[timja]

yeongsheng_tan:

Suggestion above by @scotthebert works.

[timja]

rw250155:

I have a jenkins job, which has about 20 steps.
Every step has a condition depending on a expression with jenkins variables.
About 15 of the steps call simple a unix shell script, two others system groovy scripts.
The size of the unix shell scripts vary from 5 to 30 lines.

To open this job for configure takes more than 2 minutes.
I haven't seen a difference in time if I use the workaround mentioned by Scott Hebert or not.

[timja]

chriscralle:

There is a problem with CodeMirror and textarea windows.
I just upgraded to LTS 1.642.3 this weekend and getting a little rubbing from my comrades about how long the configure load times are.
I have one 2x4 Matrix multi-step job with conditionals that have shell, python, and windows batch executes. The configure page for this job generates 8-10 timeouts in codemirror.js and takes minutes to load.
Common timeouts are on lines 1247, 1207, 853
These deal with the gutter and scrollbar as far as I can tell.

I've spent Saturday trying everything I can to somehow just DISABLE codemirror. But since it's core, not a plugin, and there's no exposed configuration for it, I'm pretty much stuck with it's behavior. Is it possible to turn it off?
The setting in security for syntax highlighting is not related, the number of builds or history is not related. This is purely codemirror barfing on complex jobs with recursive textareas.

I'm surprised codemirror is not a tag/label on this ticket.

I am creating a sanitized config.xml which I hope to upload to this ticket in the hope that it helps with a resolution.

[timja]

cmoroexpedia:

I followed the instructions from Scott (Global Security -> In the Markup Formatter section, ensure: Safe HTML is selected and Click on Disable syntax highlighting) and now I can configure jobs in Firefox, but Chrome and Safari are still broken. At least we have a workaround.

[timja]

hai_lh:

Any idea on how to turn off codemirror to bypass this issue?

[timja]

rw250155:

Hello,

it seems that the bug is fixed.
I have two jobs, which took about 10 minutes to open for edit with jenkins up to 2.11
Since I use 2.12 the same jobs takes about 10 to 20 seconds to open for edit.
And the browser does'n allocated about 10 GB of memory to open the job any more.

[timja]

danielbeck:

Resolved as JENKINS-32027 in 2.12.

[timja]

[Originally related to: JENKINS-32027]