this is because in db not actual data apply this patch and restart db-omatic diff --git a/src/db-omatic/db_omatic.rb b/src/db-omatic/db_omatic.rb index 4afffb1..c499610 100755 --- a/src/db-omatic/db_omatic.rb +++ b/src/db-omatic/db_omatic.rb @@ -74,6 +74,64 @@ class DbOmatic < Qpid::Qmf::Console domain[:synced] = true end + #find hostname from values['node'] where values['class_type'] == 'domain' + def get_host_id(abank,bbank) + begin + @cached_objects.keys.each do |objkey| + if @cached_objects[objkey][:agent_bank].to_s == abank and @cached_objects[objkey][:broker_bank].to_s == bbank and @cached_objects[objkey][:class_type].to_s == 'node' + return Host.find(:first, :conditions => ['hostname = ?',@cached_objects[objkey]["hostname"].to_s]).id + break + end + end + rescue => ex + log("error in get_host_id") + log(ex) + end + end + + def set_host(values,digit) + begin + vm = Vm.find(:first, :conditions => ['description = ?',values["name"].to_s]) + if vm and digit + vm.host_id = digit + vm.save! + else + log("this vm not exist #{values["name"]}") + end + rescue => ex + puts "error when set_host for #{values["name"]}" + puts ex + end + end + + def start_crashed_vm(vm) + task = VmTask.new( :user => 'db-omatic', :task_target => vm, :action => 'start_vm', :state => 'queued') + task.save! + log("set task for start crashed vm #{<a href="http://vm.id">vm.id</a>}") + end + + def set_domain_stopped(domain) + begin + vm = Vm.find(:first, :conditions => ['uuid = ?', domain['uuid']]) + if vm != nil + curstate = vm.state + vm.state = Vm::STATE_STOPPED + vm.host_id = nil + vm.save + domain['state'] = 'crashed' # and now i will use ipmi for reboot anavailable host - simple fencing + # if curstate == Vm::STATE_RUNNING and vm.ha # vm.ha true or false + # start_crashed_vm(vm) + # end + else + log('vm == nil ') + end + log("domain #{domain['id']} already stopped") + rescue => ex + log("can\'t set domain #{domain['id']} stopped") + log(ex) + end + end + def update_host_state(host_info, state) db_host = Host.find(:first, :conditions => [ "hostname = ?", host_info['hostname'] ]) if db_host @@ -131,6 +189,7 @@ class DbOmatic < Qpid::Qmf::Console domain_state_change = false + change_node = false obj.properties.each do |key, newval| if values[key.to_s] != newval values[key.to_s] = newval @@ -138,12 +197,30 @@ class DbOmatic < Qpid::Qmf::Console if type == "domain" and key.to_s == "state" domain_state_change = true end + if type == "domain" and key.to_s == "node" + change_node = true + end + + end end if domain_state_change update_domain_state(values) end + if change_node + values.each do |key,val| + if key == 'state' and val == 'running' + abank = values['node'].to_s.split('-')[3] + bbank = values['node'].to_s.split('-')[4] + @@host_id = get_host_id(abank,bbank) + set_host(values,@@host_id) + log("update node data for #{values['name']}") + break + end + end + end + if new_object if type == "node" @@ -187,11 +264,6 @@ class DbOmatic < Qpid::Qmf::Console end end - - def del_agent(agent) - agent_disconnected(agent) - end - # This method marks objects associated with the given agent as timed out/invalid. Called either # when the agent heartbeats out, or we get a del_agent callback. def agent_disconnected(agent) @@ -205,8 +277,10 @@ class DbOmatic < Qpid::Qmf::Console if values[:class_type] == 'node' update_host_state(values, Host::STATE_UNAVAILABLE) elsif values[:class_type] == 'domain' - update_domain_state(values, Vm::STATE_UNREACHABLE) - end + set_domain_stopped(values) + values[:timed_out] = true + @cached_objects.delete(objkey) + end end values[:timed_out] = true end @@ -248,6 +322,7 @@ class DbOmatic < Qpid::Qmf::Console db_vm = Vm.find(:all) db_vm.each do |vm| log "Marking vm #{vm.description} as stopped." + vm.host_id = nil vm.state = Vm::STATE_STOPPED vm.save end <div class="gmail_quote">2009/1/30 Hugh O. Brock <<a href="mailto:hbrock@redhat.com">hbrock@redhat.com</a>> <blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"> <div class="Ih2E3d">On Fri, Jan 30, 2009 at 01:24:36PM -0600, Carb, Brian A wrote: > date --utc on all nodes and the appliance shows that they are within a few seconds of each other. > > brian carb > unisys corporation - malvern, pa > <a href="mailto:brian.carb@unisys.com">brian.carb@unisys.com</a> > > -----Original Message----- > From: Perry Myers [mailto:<a href="mailto:pmyers@redhat.com">pmyers@redhat.com</a>] > Sent: Friday, January 30, 2009 2:19 PM > To: Carb, Brian A > Cc: <a href="mailto:ovirt-devel@redhat.com">ovirt-devel@redhat.com</a>; Ian Main > Subject: Re: [Ovirt-devel] oVirt Nodes become unavailable > > Carb, Brian A wrote: > > running oVirt 0.96 on fedora10... > > > > I noticed that after some time has elapsed (machines sitting idle overnight), the oVirt dashboard shows 2 of my 4 nodes as "unavailable(enabled)". I can access them via their consoles though, so the machines are up. Restarting the browser session does not change this. If I shutdown and then restart the oVirt server appliance, all 4 show as unavailable. Do I have to do something to make the nodes available (short of restarting them)? > > > > Any ideas? Thanks. > > Really quick thing.... check timestamps on each Node and the appliance (date --utc) > > If they're not within a few seconds of each other then kerberos gets unhappy... If that is the case then our NTP setup is not working which needs to be fixed. > > If they are in sync then Ian should be able to help > > Perry </div>Hmm... sounds like it could be a qpid or host-browser problem. Ian? --Hugh <div><div></div><div class="Wj3C7c"> _______________________________________________ Ovirt-devel mailing list <a href="mailto:Ovirt-devel@redhat.com">Ovirt-devel@redhat.com</a> <a href="https://www.redhat.com/mailman/listinfo/ovirt-devel" target="_blank">https://www.redhat.com/mailman/listinfo/ovirt-devel</a> </div></div></blockquote></div> -- С уважением, Дмитрий.