Hi, A big question that we get often asked as Ganeti team is: could you report the reason why an instance is down? In particular people would love to know whether an instance was shut down by the user from inside, crashed (and perhaps why), or was shutdown from the Xen side. It would be great if Xen could report this information somehow to a monitoring system. Currently a colleague of mine has found some of this information in the Xen logs, and also a patch floating on the list that was never merged. Is anybody coming to the Dublin Xen Hackathon in May willing to help on this (either mentoring one of us to get the change done&merged or doing the change while we implement the ganeti reporting side of it)? Thanks a lot!! Guido
On Tue, 2013-04-09 at 11:36 +0100, Guido Trotter wrote:> Hi, > > A big question that we get often asked as Ganeti team is: could you > report the reason why an instance is down? In particular people would > love to know whether an instance was shut down by the user from > inside, crashed (and perhaps why), or was shutdown from the Xen side. > It would be great if Xen could report this information somehow to a > monitoring system.Xen reports this status to the toolstack, the SHUTDOWN_* code in xen//include/public/sched.h is exposed in the domain info of the domain before it is destroyed.> > Currently a colleague of mine has found some of this information in > the Xen logs, and also a patch floating on the list that was never > merged. Is anybody coming to the Dublin Xen Hackathon in May willing > to help on this (either mentoring one of us to get the change > done&merged or doing the change while we implement the ganeti > reporting side of it)?I''d be happy to advise on a libxl/xl implementation of some way to export this info outside the toolstack -- perhaps the most generically useful way to approach this would be for xl support for calling hook scripts on various events, not just domain termination. Ian.> > Thanks a lot!! > > Guido > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xen.org > http://lists.xen.org/xen-devel
On Wed, Apr 10, 2013 at 2:30 PM, Ian Campbell <Ian.Campbell@citrix.com> wrote: Hi,> On Tue, 2013-04-09 at 11:36 +0100, Guido Trotter wrote: >> Hi, >> >> A big question that we get often asked as Ganeti team is: could you >> report the reason why an instance is down? In particular people would >> love to know whether an instance was shut down by the user from >> inside, crashed (and perhaps why), or was shutdown from the Xen side. >> It would be great if Xen could report this information somehow to a >> monitoring system. > > Xen reports this status to the toolstack, the SHUTDOWN_* code in > xen//include/public/sched.h is exposed in the domain info of the domain > before it is destroyed. > >> >> Currently a colleague of mine has found some of this information in >> the Xen logs, and also a patch floating on the list that was never >> merged. Is anybody coming to the Dublin Xen Hackathon in May willing >> to help on this (either mentoring one of us to get the change >> done&merged or doing the change while we implement the ganeti >> reporting side of it)? > > I''d be happy to advise on a libxl/xl implementation of some way to > export this info outside the toolstack -- perhaps the most generically > useful way to approach this would be for xl support for calling hook > scripts on various events, not just domain termination. >Sounds great, thanks! We''ll add it to the hackathon wiki page, then, thanks for volunteering to mentor! :) Guido