>>> On 7/18/2008 at 2:06 PM, in message <[EMAIL PROTECTED]>, Ivan <[EMAIL PROTECTED]> wrote: > Hi Robert, > > Thanks for clarifying it and providing some background. I knew that the > HASI and pretty much everything behind it was a demo for Brainshare but
Nod, that was March 2006, the code was developed initially for that demo; but then subsequently integrated, QA'd and shipped with the HASI, later in the year when SLES10 was released. We also gave a demo at LinuxWorld of the shipping product later that Fall. > that's the kind of thing we needed so bad and we use ever since (in > production for more than a year). Failures are rare but the HASI did its > job great several times ever since. Perhaps I shouldn't use it in prod > but it's still cool stuff. :) Well, it is definitely supported for production usage, as an enterprise product; that Novell runs thru various QA processes. >> It's been a while, but iirc, registered VMs still show up even if not > running; >> because they are still known/declared resources to Xen... > > Only if the VM was created by YaST GUI or some other utility perhaps > manually registered with xm new. All my HA managed VMs are clones and > because I change only the memory and disk bits in the xm config I do not > suffer from VM showing up in virtman unless its running(of course all > xenstores were cleaned at some stage). > >> Afaik today, it's still possible to start a VM without registering it, and > iirc > > Yes, That's what I have been doing and I appreciated that I wasn't > forced to use the xenstore, old school still works ;) Cool ;-) >> that's still enough also to support migration (the Xen RA was extended later >> to support live migrate issued by Heartbeat). There might be some other > > Correct. > > I can only see problem around traditional (on ocfs2 in HASI) config > files that xen cuts the support for it one day. If that happens then > something we may need to work out for xenstore. Not sure if it's > possible to actually run the xentore tdb on ocfs2 mount on all HASI > nodes? That's what we do for the config files so in theory it should > work. Somebody had similar issue in another topic recently about syncing > xenstore and I think 1 db would be the answer not the sync. Hmm, I wonder if the xend would support that, it's an interesting idea. I think others have also suggested loading Heartbeat's CIB with more even config; perhaps to the point of avoiding local config files altogether. But things should continue to work "as-designed"; plus also in general, remaining backwards compatible has been a goal too... Hth, Robert _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
