[Users] Vdsmd is respawning trying to sample NICs

[Dan Kenigsberg]

On Mon, Jun 25, 2012 at 03:15:47PM +0100, jose garcia wrote:
> On 06/25/2012 01:24 PM, Dan Kenigsberg wrote:
> >On Mon, Jun 25, 2012 at 01:15:08PM +0100, jose garcia wrote:
> >>On 06/25/2012 12:30 PM, Dan Kenigsberg wrote:
> >>>On Mon, Jun 25, 2012 at 12:11:37PM +0100, jose garcia wrote:
> >>>>On 06/25/2012 11:37 AM, Dan Kenigsberg wrote:
> >>>>>On Mon, Jun 25, 2012 at 10:57:47AM +0100, jose garcia wrote:
> >>>>>>Good monday morning,
> >>>>>>
> >>>>>>Installed Fedora 17 and tried to install the node to a 3.1 engine.
> >>>>>>
> >>>>>>I'm getting an VDS Network exception in the engine side:
> >>>>>>
> >>>>>>in /var/log/ovirt-engine/engine:
> >>>>>>
> >>>>>>2012-06-25 10:15:34,132 WARN
> >>>>>>[org.ovirt.engine.core.vdsbroker.VdsManager]
> >>>>>>(QuartzScheduler_Worker-96)
> >>>>>>ResourceManager::refreshVdsRunTimeInfo::Failed to refresh VDS , vds
> >>>>>>= 2e9929c6-bea6-11e1-bfdd-ff11f39c80eb :
> >>>>>>ovirt-node2.smb.eurotux.local, VDS Network Error, continuing.
> >>>>>>VDSNetworkException:
> >>>>>>2012-06-25 10:15:36,143 ERROR
> >>>>>>[org.ovirt.engine.core.vdsbroker.VdsManager]
> >>>>>>(QuartzScheduler_Worker-20) VDS::handleNetworkException Server
> >>>>>>failed to respond,  vds_id = 2e9929c6-bea6-11e1-bfdd-ff11f39c80eb,
> >>>>>>vds_name = ovirt-node2.smb.eurotux.local, error =
> >>>>>>VDSNetworkException:
> >>>>>>2012-06-25 10:15:36,181 INFO
> >>>>>>[org.ovirt.engine.core.bll.VdsEventListener] (pool-3-thread-49)
> >>>>>>ResourceManager::vdsNotResponding entered for Host
> >>>>>>2e9929c6-bea6-11e1-bfdd-ff11f39c80eb, 10.10.30.177
> >>>>>>2012-06-25 10:15:36,214 ERROR
> >>>>>>[org.ovirt.engine.core.bll.VdsNotRespondingTreatmentCommand]
> >>>>>>(pool-3-thread-49) [1afd4b89] Failed to run Fence script on
> >>>>>>vds:ovirt-node2.smb.eurotux.local, VMs moved to UnKnown instead.
> >>>>>>
> >>>>>>While in the node, vdsmd does fail to sample nics:
> >>>>>>
> >>>>>>in /var/log/vdsm/vdsm.log:
> >>>>>>
> >>>>>>    nf = netinfo.NetInfo()
> >>>>>>   File "/usr/share/vdsm/netinfo.py", line 268, in __init__
> >>>>>>     _netinfo = get()
> >>>>>>   File "/usr/share/vdsm/netinfo.py", line 220, in get
> >>>>>>     for nic in nics() ])
> >>>>>>KeyError: 'p36p1'
> >>>>>>
> >>>>>>MainThread::INFO::2012-06-25 10:45:09,110::vdsm::76::vds::(run) VDSM
> >>>>>>main thread ended. Waiting for 1 other threads...
> >>>>>>MainThread::INFO::2012-06-25 10:45:09,111::vdsm::79::vds::(run)
> >>>>>><_MainThread(MainThread, started 140567823243072)>
> >>>>>>MainThread::INFO::2012-06-25 10:45:09,111::vdsm::79::vds::(run)
> >>>>>><Thread(libvirtEventLoop, started daemon 140567752681216)>
> >>>>>>
> >>>>>>in /etc/var/log/messages there is a lot of vdsmd died too quickly:
> >>>>>>
> >>>>>>Jun 25 10:45:08 ovirt-node2 respawn: slave '/usr/share/vdsm/vdsm'
> >>>>>>died too quickly, respawning slave
> >>>>>>Jun 25 10:45:08 ovirt-node2 respawn: slave '/usr/share/vdsm/vdsm'
> >>>>>>died too quickly, respawning slave
> >>>>>>Jun 25 10:45:09 ovirt-node2 respawn: slave '/usr/share/vdsm/vdsm'
> >>>>>>died too quickly for more than 30 seconds, master sleeping for 900
> >>>>>>seconds
> >>>>>>
> >>>>>>I don't know why Fedora 17 calls p36p1 to what was eth0 in Fedora
> >>>>>>16, but tried to configure a bridge ovirtmgmt and the only
> >>>>>>difference is that KeyError becomes 'ovirtmgmt'.
> >>>>>The nic renaming may have happened due to biosdevname. Do you have it
> >>>>>installed? Does any of the /etc/sysconfig/network-scripts/ifcfg-* refer
> >>>>>to an old nic name?
> >>>>>
> >>>>>Which version of vdsm are you running? It seems that it is
> >>>>>pre-v4.9.4-61-g24f8627 which is too old for f17 to run - the output of
> >>>>>ifconfig has changed. Please retry with latest beta version
> >>>>>https://koji.fedoraproject.org/koji/buildinfo?buildID=327015
> >>>>>
> >>>>>If the problem persists, could you run vdsm manually, with
> >>>>># su - vdsm -s /bin/bash
> >>>>># cd /usr/share/vdsm
> >>>>># ./vdsm
> >>>>>maybe it would give a hint about the crash.
> >>>>>
> >>>>>regards,
> >>>>>Dan.
> >>>>Well, thank you. I have updated Vdsm to version 4.10. Now the
> >>>>problem is with SSL and XMLRPC.
> >>>>
> >>>>This is the error in the side of the engine:
> >>>>
> >>>>/var/log/ovirt-engine/engine.log
> >>>>
> >>>>ERROR [org.ovirt.engine.core.vdsbroker.vdsbroker.VdsBrokerCommand]
> >>>>(QuartzScheduler_Worker-52) XML RPC error in command
> >>>>GetCapabilitiesVDS ( Vds: ovirt-node2.smb.eurotux.local ), the error
> >>>>was: java.util.concurrent.ExecutionException:
> >>>>java.lang.reflect.InvocationTargetException,
> >>>>NoHttpResponseException: The server 10.10.30.177 failed to respond.
> >>>>
> >>>>In the side of the node, there seems to be an authentication problem:
> >>>>
> >>>>/var/log/vdsm/vdsm.log
> >>>>
> >>>>SSLError: [Errno 1] _ssl.c:504: error:1407609C:SSL
> >>>>routines:SSL23_GET_CLIENT_HELLO:http request
> >>>>Thread-810::ERROR::2012-06-25
> >>>>12:02:46,351::SecureXMLRPCServer::73::root::(handle_error) client
> >>>>('10.10.30.101', 58605)
> >>>>Traceback (most recent call last):
> >>>>   File "/usr/lib64/python2.7/SocketServer.py", line 582, in
> >>>>process_request_thread
> >>>>     self.finish_request(request, client_address)
> >>>>   File
> >>>>"/usr/lib/python2.7/site-packages/vdsm/SecureXMLRPCServer.py", line
> >>>>66, in finish_request
> >>>>     request.do_handshake()
> >>>>   File "/usr/lib64/python2.7/ssl.py", line 305, in do_handshake
> >>>>     self._sslobj.do_handshake()
> >>>>SSLError: [Errno 1] _ssl.c:504: error:1407609C:SSL
> >>>>routines:SSL23_GET_CLIENT_HELLO:http request
> >>>>
> >>>>In /var/log/messages there is an:
> >>>>
> >>>>vdsm [5834]: vdsm root ERROR client ()
> >>>>
> >>>>with the ip address of the engine.
> >>>Hmm... Do you have ssl=true in your /etc/vdsm/vdsm.conf ?
> >>>Does vdsm respond locally to
> >>>
> >>>     vdsClient -s 0 getVdsCaps
> >>>
> >>>(Maybe your local certificates and key were corrupted, and you will have
> >>>to re-install the host form Engine in order to create a new set)
> >>>
> >>I have recreated the db and run engine-setup again. I have tried
> >>with ssl= true commented and uncommented in the node. vdsClient -s 0
> >>getVdsCaps works locally and provides the information of the host,
> >>but something seems to be preventing it to get to the engine. I am
> >>still getting the same error. The installer is not beginning. I can
> >>do ssh as root to the host and vdsmd is alive.
> >What do you mean by "The installer is not beginning"?
> >Could you review your /etc/pki/vdsm/certs/ and check that they have been
> >generated by *your* engine? Is the cacert the same as the one on the
> >Engine machine?
> >
> >Dan.
> In the engine server I have a self signed certificate. I set up a
> cacert and server.pem to avoid libvirtd complaining about gssapi.
> The one in the node is issued by the VDSM Certificate Authority, so
> I suppose it is set up by the vdsm package installation.

So here lies the problem. When Vdsm is first installed, it generates its
own self-signed certificate. This, by definition, does not help to
identify it for Engine.

When you add a host to a data center, Engine logs into the host and
askes the host to produce a *new* key. Engine's CA then signs the key,
and put the cert back under /etc/pki/vdsm/cert.

Something has gone wrong in this process. If you re-install the host it
*should* override current keys. If not - it is a bug. Could you look at
the installation logs (they sit on a random dir under /tmp)?
Maybe there's a clue there why you keep the default
good-for-almost-nothing keys that come with vdsm.

Dan.
>