On 12/07/2011 08:17 AM, Dennis Hübner wrote:
Am 07.12.11 14:09, schrieb Denis Roy:I guess that's worse than Hudson over NFS, which is fragile but fast? :-)No, that not right. Hudson access over https is fragile and slow.Well, then we have a problem. /shared is a NFS-mounted filesystem which is shared amongst all the build/hudson servers. The NFS mount is problematic since you can't clean your workspace.So we *moved /shared/jobs* to a locally-mounted filesystem, which is *not shared*. As far as I understand it, the new location shouldn't matter as you can use the https:// links to refer to the build artifacts. Right? I understand... But what are the options? A shared filesystem isn't exactly working right now.It's not just a URL that have to be changed. We talk about protocol change from file: to http/-s: !?You will need to update your build/promote scripts to no longer refer to /shared/jobs. /opt/users/hudsonbuild... is the local working directory for each master/slave node. That directory has never been shared, so if you refer to files in that path, your script may break if the job runs elsewhere.No it's file:/optBTW: file:/opt/users/hudsonbuild/ shouldn't that be: file:///opt/users/hudsonbuild... ? Unless I am mistaken, the most reliable way to refer to build artifacts, regardless of which slave/master you're using is http/s. Please correct me if I am wrong. As I understand it, most Hudson installs employ slaves which do not have a shared (amongst themselves) filesystem. Denis Denis On 12/07/2011 07:17 AM, Sven Efftinge wrote:Xtext and MWE are going to have a release today. But the problems Dennis mentioned are holding us back. Sven On Dec 7, 2011, at 11:23 AM, Dennis Hübner wrote:Hudson links aka file:/opt/users/hudsonbuild/.hudson/jobs/MWE-nightly-HEAD/lastSuccessful/archive/mwe.p2.repository/ doesn't work either Means we can't fetch latest artifacts from jobs we depend on. Regards, Dennis. Am 07.12.11 10:58, schrieb Dennis Hübner:Hello Matt, I'm not quite sure that my issue is related to the latest changes, but I just noticed that our lastSuccessful/lastStable file system links a not up to date. Hudson web link https://hudson.eclipse.org/hudson/job/MWE-nightly-HEAD/lastStableBuild/ points to 07.12.2011 build and file system link under /shared/jobs/MWE-nightly-HEAD/ still points to the old stable: lrwxrwxrwx 1 hudsonBuild callisto-dev 26 2. Dez 13:11 lastStable -> builds/2011-12-02_13-04-30 Our promote scripts are using file system URL /shared/jobs/MWE-nightly-HEAD/ so right now I can't promote! Regards, Dennis. Am 06.12.11 17:46, schrieb Webmaster(Matt Ward):Hi Folks, As some of you know we've been having some issues recently with Hudson not letting go of files which causes issues when people try to clean up their workspace. Denis and I have come up with a potential fix that will allow us to bypass NFS by mounting a disk image locally on the master that will contain the 'workspace' data. However I need to copy the current data over and then restart Hudson. While short notice my plan is to do this today starting at 3pm. I expect Hudson will be offline for about 20 minutes while the data syncs and I re-configure the master. if you absolutely must get your M build done and this will prevent that, please let me know ASAP. -Matt.-- Dennis Hübner Softwareentwickler Mobil: 0151 173 96 707 http://www.itemis.de/ itemis AG Am Germaniahafen 1 24143 Kiel Rechtlicher Hinweis: Amtsgericht Dortmund, HRB 20621 Vorstand: Wolfgang Neuhaus, Jens Wagener, Dr. Georg Pietrek Aufsichtsrat: Prof. Dr. Burkhard Igel (Vors.), Stephan Grollmann, Michael Neuhaus _______________________________________________ --
Denis Roy Director, IT Services Eclipse Foundation, Inc. -- http://www.eclipse.org/ Office: 613.224.9461 x224 (Eastern time) denis....@eclipse.org  |
_______________________________________________ cross-project-issues-dev mailing list cross-project-issues-dev@eclipse.org https://dev.eclipse.org/mailman/listinfo/cross-project-issues-dev
