OSDir


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: advanced networking with public IPs direct to VMs


Update to this.


I ran the all on one NIC test again and it does report as "Down" in the UI as opposed to "Alert" when using multiple NICs.


Looking at the management server log this seems to be the key part -


1) from the single NIC logs -


2018-06-04 10:17:10,967 DEBUG [c.c.h.KVMInvestigator] (AgentTaskPool-3:ctx-8627b348) (logid:ef7b8230) Neighbouring host:5 returned status:Down for the investigated host:4
2018-06-04 10:17:10,967 DEBUG [c.c.h.KVMInvestigator] (AgentTaskPool-3:ctx-8627b348) (logid:ef7b8230) HA: HOST is ineligible legacy state Down for host 4
2018-06-04 10:17:10,967 DEBUG [c.c.h.HighAvailabilityManagerImpl] (AgentTaskPool-3:ctx-8627b348) (logid:ef7b8230) KVMInvestigator was able to determine host 4 is in Down
2018-06-04 10:17:10,967 INFO  [c.c.a.m.AgentManagerImpl] (AgentTaskPool-3:ctx-8627b348) (logid:ef7b8230) The agent from host 4 state determined is Down
2018-06-04 10:17:10,967 ERROR [c.c.a.m.AgentManagerImpl] (AgentTaskPool-3:ctx-8627b348) (logid:ef7b8230) Host is down: 4-dcp-cscn2.local. Starting HA on the VMs



2) from the setup with 2 NICs (managemnet/storage on one NIC, guest traffic on the other) -



2018-06-04 12:53:45,853 WARN  [c.c.h.KVMInvestigator] (AgentTaskPool-3:ctx-0aed2673) (logid:32aaef2a) Agent investigation was requested on host Host[-2-Routing], but host does not support investigation because it has no NFS storage. Skipping investigation.
2018-06-04 12:53:45,854 DEBUG [c.c.h.HighAvailabilityManagerImpl] (AgentTaskPool-3:ctx-0aed2673) (logid:32aaef2a) KVMInvestigator was able to determine host 2 is in Disconnected
2018-06-04 12:53:45,854 INFO  [c.c.a.m.AgentManagerImpl] (AgentTaskPool-3:ctx-0aed2673) (logid:32aaef2a) The agent from host 2 state determined is Disconnected
2018-06-04 12:53:45,854 WARN  [c.c.a.m.AgentManagerImpl] (AgentTaskPool-3:ctx-0aed2673) (logid:32aaef2a) Agent is disconnected but the host is still up: 2-dcp-cscn2.local
2018-06-04 12:53:45,854 WARN  [o.a.c.alerts] (AgentTaskPool-3:ctx-0aed2673) (logid:32aaef2a) AlertType:: 7 | dataCenterId:: 1 | podId:: 1 | clusterId:: null | message:: Host disconnected, name: dcp-cscn2.local (id:2), availability zone: dcp1, pod: dcpp1
2018-06-04 12:53:45,858 INFO  [c.c.a.m.AgentManagerImpl] (AgentTaskPool-3:ctx-0aed2673) (logid:32aaef2a) Host 2 is disconnecting with event AgentDisconnected
2018-06-04 12:53:45,858 DEBUG [c.c.a.m.AgentManagerImpl] (AgentTaskPool-3:ctx-0aed2673) (logid:32aaef2a) The next status of agent 2is Alert, current status is Up
2018-06-04 12:53:45,858 DEBUG [c.c.a.m.AgentManagerImpl] (AgentTaskPool-3:ctx-0aed2673) (logid:32aaef2a) Deregistering link for 2 with state Alert
2018-06-04 12:53:45,859 DEBUG [c.c.a.m.AgentManagerImpl] (AgentTaskPool-3:ctx-0aed2673) (logid:32aaef2a) Remove Agent : 2


I don't know what it means by host has no NFS storage but you can see it never marks the failed node as down.


Any ideas ?




________________________________
From: Rafael Weingärtner <rafaelweingartner@xxxxxxxxx>
Sent: 04 June 2018 21:15
To: users
Subject: Re: advanced networking with public IPs direct to VMs

Everything seems to be normal at a first glance. Do you see some sort of
error in the log files?

On Mon, Jun 4, 2018 at 11:39 AM, Jon Marshall <jms.123@xxxxxxxxxxxxx> wrote:

> CS version 4.11
>
> VM HA at the moment (not Host HA as yet)
>
> KVM
>
>
> For the management node just one NIC - 172.30.3.2/26 assigned to physical
> NIC.
>
>
> For the compute nodes -
>
>
> 3 NICs so as an example from one compute node -
>
>
> ifcfg-eth0
>
> BRIDGE=cloudbr0
>
>
> ifcfg-eth1
>
> BRIDGE=cloudbr1
>
>
> ifcfg-eth2
>
> BRIDGE=cloudbr2
>
>
> then the 3 bridges -
>
>
> ifcfg-cloudbr0
>
> ip address 172.30.3.3/26    <--- management network
>
>
> if-cloudbr1
>
> ip address 172.30.4.3/25  <-- guest traffic
>
> gateway 172.30.4.1
>
>
>
> ifcfg-cloubr2
>
> ip address 172.30.5.3 /28 <-- storage traffic
>
>
> traffic labels would be cloudbr0, cloudbr1, cloudbr2
>
>
> Can only get failover working when I put all traffic on same NIC.
>
>
>
> ________________________________
> From: Rafael Weingärtner <rafaelweingartner@xxxxxxxxx>
> Sent: 04 June 2018 15:31
> To: users
> Subject: Re: advanced networking with public IPs direct to VMs
>
> What type of failover are you talking about?
> What version are you using?
> What hypervisor are you using?
> How are you configuring your NICs in the hypervisor?
> How are you configuring the traffic labels in ACS?
>
> On Mon, Jun 4, 2018 at 11:29 AM, Jon Marshall <jms.123@xxxxxxxxxxxxx>
> wrote:
>
> > Hi all
> >
> >
> > I am close to giving up on basic networking as I just cannot get failover
> > working with multiple NICs (I am not even sure it is supported).
> >
> >
> > What I would like is to use 3 NICs for management, storage and guest
> > traffic. I would like to assign public IPs direct to the VMs which is
> why I
> > originally chose basic.
> >
> >
> > If I switch to advanced networking do I just configure a guest VM with
> > public IPs on one NIC and not both with the public traffic -
> >
> >
> > would this work ?
> >
>
>
>
> --
> Rafael Weingärtner
>



--
Rafael Weingärtner