Hi Laurent, Many many thanks for this. We had the exact same problem but using xenserver as hosts. The fix for us was:
select name,broadcast_uri from networks where mode='Dhcp'; We were using basic networking as well. We had upgraded from 4.3 to 4.4.2, using Xenserver 6.1. Thank you!!!!!!! On Mon, Feb 23, 2015 at 12:40 PM, Laurent Steff <laurent.st...@inria.fr> wrote: > Hi, > > This mail to share a fight we had at INRIA upgrading our Cloudstack/KVM > farm > from 4.2 to 4.4.2 following this documentation : > > > http://cloudstack-release-notes.readthedocs.org/en/latest/upgrade/upgrade-4.2.html > > It's now solved, but I would like to share, as I think : > > - it could helps other people like us who have already migrated from > Cloudstack 3.X to 4.X > - there is one bug marked as fixed and it should not > https://issues.apache.org/jira/browse/CLOUDSTACK-7399 > - a little documentation is missing (how to test if we have the good > qemu-kvm version for systemVMs templates) > > Here are the (long) details > > Technical informations : > ------------------------ > > - Upgrade from Cloudstack 4.2.1 to 4.4.2 > - CentOS 6/KVM for agents > - official Cloudstack rpms > - 1 zone with BasicNetworking > > We are using cloudstack here in two environnments : > > - qualification, with MS and agents created on 4.2.1 > - production, with MS and agents originally created on 3.x version, long > time ago before > Apache :D > > > Qualification troubles and solution : > ------------------------------------- > > - systemVM do not start after cloudstack-sysvmadm launch > - Solution was tu upgrade the KVM agents from Centos 6.3 to 6.6 > - we think (not sure) that we had a trouble with an historical qemu-kvm > version, and a good test > to document may be : what version of CentOS qemu-kvm supports, launching > this command : > --- > /usr/libexec/qemu-kvm -M ? > --- > > > Production troubles and solution : > ---------------------------------- > > - cloudstack-sysvmadm takes hours to shutdown, upgrade and restart > systemVM (2 or 3 hours) > - starting/stopping existing instances works > - but we're unable to create new instances (error on MS : > --- > com.cloud.exception.AgentUnavailableException: Resource [Host:xx] is > unreachable: Host xx: Unable to start > instance due to Unable to get answer that is of class > com.cloud.agent.api.StartAnswer > --- > - when destroyed manually, systemVM won't restart > - debug on agents shows the same message as this bug : > https://issues.apache.org/jira/browse/CLOUDSTACK-7399 > which is officially resolved in 4.4.1 (our version is 4.4.2 !!!) > --- > WARN [cloud.agent.Agent] (agentRequest-Handler-2:null) Caught: > java.lang.NullPointerException > at > com.cloud.network.Networks$BroadcastDomainType.getSchemeValue(Networks.java:159) > ... > DEBUG [cloud.agent.Agent] (agentRequest-Handler-2:null) Seq > 25-6233544834234187813: { Ans: , MgmtId: 345044038925, via: 25, Ver: v1, > Flags: 10, > [{"com.cloud.agent.api.Answer":{"result":false,"details":"java.lang.NullPointerException\n\tat > com.cloud.network.Networks$BroadcastDomainType.getSchemeValue(Networks.java:159)\n\tat > com.cloud.network.Networks$BroadcastDomainType.getValue(Networks.java:213)\n\tat > com.cloud.hypervisor. > ... > --- > - we had to find our bascicnetwork in mysql table networks, whom > broadcast_uri was NULL > - and modify it to the "new" style vlan://untagged : > --- > update networks set broadcast_uri="vlan://untagged" where id="our > bascinetwork id"; > > Hope it could help, > > -- > Laurent Steff > > DSI/SESI > INRIA > http://www.inria.fr/ >