Is it possible to have a gerrit gating criteria on a periodic stable build as well?
If there are some "slow" tests that can run once/twice a day (and not on every patch), we may be able to have a periodic run which once it reaches to a certain stable level it can become gating (merges will be permitted only if it's not broken) - this methodology is kind of a "find the break after patch is merged" but finding the breaking merge and reverting it can be pretty easy. Thanks, Gal -----Original Message----- From: [email protected] [mailto:[email protected]] On Behalf Of Andrew Grimberg Sent: Thursday, July 07, 2016 12:31 AM To: Robert Varga <[email protected]>; Jamo Luhrsen <[email protected]>; Anil Vishnoi <[email protected]>; Josh Hershberg <[email protected]> Cc: Jamo Luhrsen <[email protected]>; Peer, Livnat <[email protected]>; [email protected]; [email protected]; [email protected]; [email protected] Subject: Re: [integration-dev] [opendaylight-dev] [releng] [openflowplugin-dev] More automated testing, please! And make it gating. On 07/06/2016 12:59 PM, Robert Varga wrote: > On 07/06/2016 07:28 PM, Andrew Grimberg wrote: >> Secondly, committers would _lose_ the verify rights as they would >> become completely managed by Jenkins. The values would change from -1 >> .. +1 to become -1 .. +2 >> >> Thirdly, committers would _lose_ the submit right. This would be >> completely controlled by Jenkins based upon state transitions. >> Effectively turning the "Approve +1" right that committers would be >> gaining into what would effectively be a "conditionally submit this >> code, provided all tests pass" operation. > > While I like the overall idea, I certainly do not have enough faith in > infrastructure and verification pipeline stability. > > I presume those test definitions will be in JJB, which means that when > infra goes south (as it has numerous times over the past year) or a > project goes AWOL (and looses its snapshot artifacts, for example), we > are dead in the water with no options to move forward until whatever > went wrong gets resolved -- which is a heavily PST-centric process and > has taken days (and weeks) to complete on more than one occasion. > > I am sorry, but this proposal will need to show a serious long-term > SLA before committers can be *asked* to relinquish these rights. Migration to this system would be by project election. The whole point is that folks are asking for CSIT based merge gating. To make the gating operate cleanly we have to move to certain functions to being managed by the CI system itself. So we're not making a blanket ask for committers to relinquish rights, it will be voluntary. I'll also point out that we have a new APAC based release engineering coming up to speed on our infrastructure so it's not going to be 100% North America time resolution of issues anymore. -Andy- -- Andrew J Grimberg Systems Administrator Release Engineering Team Lead The Linux Foundation _______________________________________________ openflowplugin-dev mailing list [email protected] https://lists.opendaylight.org/mailman/listinfo/openflowplugin-dev
