So far, rolling back the plugin version from 1.1.26 to 1.1.21 looks promising. We were seeing the issue at least once a day, and after 24 hours it hasn't recurred. In fact, a job which has been blocked for a week on this issue (we just didn't notice because it isn't a critical job and only runs once daily) ran automatically this morning.
Perhaps my issue isn't the same as yours, but it was definitely caused by some change in git plug-in behavior between 1.1.21 and 1.1.26. I had never seen calls to 'git show' before the upgrade, so I had assumed that the 'git show' command was the new behavior. Based on what you are telling me, though, it sounds like the parameters to git show changed, or perhaps the way the output is parsed... What version of git are you using? I am on 1.7.10. Maybe that is the difference? On Thu, Feb 14, 2013 at 9:41 PM, Mark Waite <[email protected]> wrote: > I am interested if the rollback of the git plugin version will change the > behavior you are observing. I'm pretty sure a plugin version rollback > would not have resolved my issue, since my issue was due to a bug in the > "git show" command and a change in my git repository which exercised that > bug. > > If you have excluded regions, or excluded users, then the submission I > referenced is unlikely to help you. The submission helps me because I have > very few jobs that use excluded regions or users. Most of the time the new > git plugin will entirely avoid the call to "git show" in my usage. If you > have excluded regions or excluded users, I don't know of any way to avoid > the calls to "git show", which probably means you will continue to see the > problem. > > The issue I encountered was that the repository that Jenkins was using > seems to exercise a known bug in the "git show" command. It seems that > there are cases where the "git show" command may spin with 100% CPU use for > hours. I realize that is not the problem you encountered, but it was the > original problem I encountered which caused me to investigate why the git > plugin was calling "git show" at all. > > The plugin calls "git show" within the "isRevExcluded()" method. It uses > "git show" to decide if a specific commit has been excluded or not. > Unfortunately, by using "git show", it also exercised the bug in "git > show" which exposed the initial problem I saw. > > Thanks, > Mark Waite > > ------------------------------ > *From:* Don Ross <[email protected]> > *To:* [email protected] > *Sent:* Thursday, February 14, 2013 8:40 AM > > *Subject:* Re: Support for git plug-in? > > Thanks for the info, Mark. I will roll my installation back to the > previous version (1.1.21) and wait for an update. I am a little concerned > because we do, in fact, have exclusion regions, and have since before we > started using Git. Does that mean that this fix (when available) may not > actually solve my problem? > > My company did have CloudBees out to give a sales pitch on Enterprise > Jenkins this fall, but decided against it. > > On Wed, Feb 13, 2013 at 1:27 PM, Mark Waite <[email protected]> wrote: > > I concur with Stephen and Nicolas about this being a community effort and > commend them for their contribution and for the help of their employer. > > I think the community may have already discovered a variant this problem > and has already provided a fix for some of the cases. That fix has already > been pulled into the source code of the git-plugin through a community > contribution. It has not yet been released in a new version of the git > plugin. > > If your git configuration for this job does not define any exclusion > regionss, does not define any excluded committers, and does not define any > inclusion regions, then the next release of the git plugin (whenever that > happens) will resolve your issue. > > Refer to > https://github.com/jenkinsci/git-plugin/commit/81f432ef0961149b13ff82d7803a897587d8c887for > the details of the commit. > > Thanks, > Mark Waite > > I agree with Stephen and Nicolas that enterprise support is available from > their employer (I suspect that is CloudBees). > > There were a number of interesting discussions about the inherent problems > in the "git show" command and its use in the git plugin. As far as I can > tell, this is the one and only use of the "git show" command, and it can be > avoided in most cases (with the change mentioned above). > > ------------------------------ > *From:* Stephen Connolly <[email protected]> > *To:* "[email protected]" <[email protected]> > *Sent:* Wednesday, February 13, 2013 7:18 AM > *Subject:* Re: Support for git plug-in? > > On 13 February 2013 13:36, Don Ross <[email protected]> wrote: > > I am sorry, this is my first foray into trying to get support for a > community-maintained tool. We are using Jenkins in an enterprise > environment > > > [My & Nicolas' employer's hat] > > If you were/are a paying customer of My & Nicolas' employer's you > could/can apply whips and chains to Nicolas to get the bug fixed > > (Nicolas, no cheating and assigning to me ;-) ) > > [/My & Nicolas' employer's hat] > > Sorry, I couldn't help myself there... I did remove the name of My & > Nicolas' employer, though it isn't exactly a secret. > > > and haven't had any issues we couldn't solve ourselves since I started > working here. Thanks for getting back to me directly. > > My issue is reported as https://issues.jenkins-ci.org/browse/JENKINS-16693; > basically, we are getting heap errors since upgrading to the latest plug-in > version, but only in our production environment. We can't reproduce it in > the supposedly-identical test environment or we wouldn't have upgraded. I > think it is due to the enhancements made in version 1.1.23 to changelog > parsing for merge targets, but that is just speculation on my part. > > Any debugging suggestions you have would be welcome. > > Thanks, > Don > > > On Wed, Feb 13, 2013 at 8:21 AM, nicolas de loof <[email protected] > > wrote: > > yes it is, but filling a bug (which one ?) doesn't mean someone will jump > on it to fix. This is a community effort :P > issues are by default assigned to plugin maintainer (me) so the status on > jira. I don't really monitor it (my bad), as I mostly focus on my own > issues with this plugin and contributors pull-requests. > > 2013/2/13 <[email protected]> > > Is the Git plug-in still supported? I filed a bug issues.jenkins-ci.orglast > week and haven't seen any movement on it, or on any of the issues > filed in the last six months against the Git component. There are about > fifty such issues, most of them assigned to Nicholas de Loof (the original > author, I think), and none of them with any response or acknowledgement > from him (or anyone else). > > -- > You received this message because you are subscribed to the Google Groups > "Jenkins Developers" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/groups/opt_out. > > > > > -- > You received this message because you are subscribed to the Google Groups > "Jenkins Developers" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/groups/opt_out. > > > > > -- > You received this message because you are subscribed to the Google Groups > "Jenkins Developers" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/groups/opt_out. > > > > > -- > You received this message because you are subscribed to the Google Groups > "Jenkins Developers" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/groups/opt_out. > > > > > -- > You received this message because you are subscribed to the Google Groups > "Jenkins Developers" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/groups/opt_out. > > > > > -- > You received this message because you are subscribed to the Google Groups > "Jenkins Developers" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/groups/opt_out. > > > > > -- > You received this message because you are subscribed to the Google Groups > "Jenkins Developers" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > For more options, visit https://groups.google.com/groups/opt_out. > > > -- You received this message because you are subscribed to the Google Groups "Jenkins Developers" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/groups/opt_out.
