Created attachment 1256041 [details] All logs from host Description of problem: VM can not obtain dhcp IP after migrate from el6-3.5-vintage to el7-3.6-ngn Version-Release number of selected component (if applicable): From: rhev-hypervisor6-6.6-20150128.0.iso To: RHVH-3.6-20170217.5-RHVH-x86_64-dvd1.iso How reproducible: 100% Steps to Reproduce: 1. Install el6-3.5-vintage rhev-hypervisor6-6.6-20150128.0.iso, add el6-3.5-vintage to engine 3.6(with 3.5 cluster), setup nfs storage to host, and create vm on host, vm can get dhcp IP. Below are detailed steps: Original env: el6-3.5-vintage + engine3.6(3.5 cluster) ================ 1.1 Installe rhev-hypervisor6-6.6-20150128.0.iso 1.2 Registee host into RHEVM 3.6 1.3 Adde data storage (NFS) 1.4 Adde ISO storage (NFS) 1.5 Create disk 1.6 Create virtual machine (disk attached), set network as dhcp mode, can get dhcp IP 1.7 Before the migration, change the interval of Ovf update task to make sure all vms will be in the storage. This task is usually executed each 60minutes. # engine-config -s OvfUpdateIntervalInMinutes=1 # service ovirt-engine restart (Wait at least 1 minute after restart) 2. Set maintenance of host el6-3.5-vintage in engine 3.6, and delete host(with DC and cluster) from engine 3.6 side. 3. Re-install RHVH-3.6-20170217.5-RHVH-x86_64-dvd1.iso in the same host, add it to engine 3.6(3.6 cluster), migrate vm to RHVH-3.6. Below are detailed steps: On el7-3.6-ngn + engine3.6(3.6 cluster) side: =============== 3.1 Initial Setup for RHVH and RHEVM: - Install el7-3.6-ngn - Registered host into RHEVM 3.6 - Enable different NFS Storage as data domain in the datacenter, just to make it UP. 3.2 Now import the NFS Storage from RHEVM: - Storage tab -> Import Domain -> Provide the NFS Storage from step 1.3 3.3 In the Datacenter tab, select storage and active the imported domain 3.4 Import the VM/Disk: - Storage tab -> Select the domain imported -> Select VM Import subtab -> Select VM and Click import 3.5 After import VM, run VM, check dhcp IP Actual results: After step 3.5, VM can not get dhcp IP Expected results: After step 3.5, VM should obtain dhcp IP successful, just same as step 1.6 Additional info:
Created attachment 1256090 [details] logs from engine
Can you please attach an sosreport? Can VMs created on a 3.6 NGN install (without migrating from 3.5) receive IPs?
(In reply to Ryan Barry from comment #2) > Can you please attach an sosreport? I added it to attachment "All logs from host": /var/log/sosreport-dhcp-10-16.nay.redhat.com-20170221072723.tar.xz > > Can VMs created on a 3.6 NGN install (without migrating from 3.5) receive > IPs? Yes, it can receive IP. I will send ENV info to you via email, please access it if needed.
Huijuan, could you use the last el6-3.5-vintage for upgrade testing for bug 1421098 and this bug? In bug description, the rhev-hypervisor6-6.6-20150128.0 was vintage 3.5.0 GA version which was version 2 years ago, the version is not ideal for upgrade testing, although not sure it related to this bug.
OK, Ying. We tested with latest el6-3.5-vintage build rhev-hypervisor6-6.8-20160707.3.iso, also encountered this bug. And we will test Bug 1421098 with the latest el6-3.5-vintage build from now on. Thanks.
I miss part of the process: in what stage was the data storage domain detached from Engine? Does a fresh VM started on the el7-3.6-ng host get dhcp properly? What happens if the imported VM is run on a different host?
Yevgeni, can you spot something broken in the import process?
Huijuan Zhao, 1. What's the network the VM was connecting a DHCP sever prior the upgrade? 2. Is that network present and is plugged on the VM after the upgrade and import? 3. If 2 is yes, is the network in the new DC identically configured (vlan, mtu, etc.) as the one in the DC prior the upgrade.
(In reply to Yevgeny Zaspitsky from comment #8) > Huijuan Zhao, > > 1. What's the network the VM was connecting a DHCP sever prior the upgrade? The nic is chosen with the type VirtIO > 2. Is that network present and is plugged on the VM after the upgrade and > import? The network is still plugged and imported. > 3. If 2 is yes, is the network in the new DC identically configured (vlan, > mtu, etc.) as the one in the DC prior the upgrade. It only import VM, and no more steps, nic is still here but VM can not obtain dhcp IP after migrate from el6-3.5-vintage to el7-3.6-ngn Jian Wu Thanks
(In reply to Dan Kenigsberg from comment #6) > I miss part of the process: in what stage was the data storage domain > detached from Engine? Sorry, maybe I cannot fully understand what did you mean, the storage domain was created with nfs at begin, and just imported when new host is up, keep the nfs not to be destroyed. > Does a fresh VM started on the el7-3.6-ng host get dhcp properly? Yes, it could get dhcp when a fresh VM started on the el7-3.6-ng host > What happens if the imported VM is run on a different host? I think you mean that VM should happen when is imported into another host, VM will work normally like a fresh VM created. Jian Wu Thanks
JianWu, the process described in this bug is very complex. I am not sure I completely understand what you are doing, and I suspect that you are using confusing terms (e.g. you never import a VM into a host; VMs can be imported to a datacenter or an Engine). Upgrading a host to el7-3.6-ng has very little to do with importing a storage domain to ovirt-engine. Why are you testing a flow that includes both actions? Can we split it instead to two flows, one of detaching/reattching the NFS storage domain, the other is of host upgrade? Could you describe your setup completely? How many hosts do you have there? How many storage domains? When exactly do you dettach the storage domain that you later import?
(In reply to Dan Kenigsberg from comment #11) > JianWu, the process described in this bug is very complex. I am not sure I > completely understand what you are doing, and I suspect that you are using > confusing terms (e.g. you never import a VM into a host; VMs can be imported > to a datacenter or an Engine). > > Upgrading a host to el7-3.6-ng has very little to do with importing a > storage domain to ovirt-engine. Why are you testing a flow that includes > both actions? Can we split it instead to two flows, one of > detaching/reattching the NFS storage domain, the other is of host upgrade? Sure, the fact is that there was a bug about migrate VM from original host to another host as customers' need, so we take these step to finish this aim. el7-3.6-ng is a build that we hope to function well on this scenario migrating VM from el6-3.5-vintage > > Could you describe your setup completely? How many hosts do you have there? > How many storage domains? When exactly do you dettach the storage domain > that you later import? In fact, our aim is to migrate VM from original host to another host, so at first, we add host 1 to engine, set nfs storage as storage domain and iso storage with type "nfs", then create a VM under this Datacenter secondly, we install a new host 2 and register to the proper engine, and add a nfs storage domain, then import domain provided by host1 so that we could keep the original VM be used in new host 2. Upon this, the VM could continue to work even we change to a new host.
Thanks Dan Kenigsberg, Douglas, jiawu, cshao and ycui to clarify the testing steps, with the below steps, no such issue any more. So I will close this bug. Test Versions: RHEV Hypervisor - 6.8 - 20160707.3.el6ev rhvh-3.6-0.20170307.0 Test steps: 1. Install rhev-hypervisor6-6.8-20160707.3.iso on host 1 2. In RHEVM 3.6, add host 1(el6-3.5-vintage) to RHEVM 3.6 (3.5 cluster), add NFS data/ISO storage, create VM on host 1. Below are detail steps: -2.1 Create Data Center with 3.5 compatibility version -2.2 In networks tab, create a new network named "rhevm" -2.3 Create cluster35(3.5 Compatibility Version), using "rhevm" as Management Network Note: Select the Management Network as "rhevm" -2.4 Add host 1 (rhev-hypervisor6-6.8-20160707.3) to this cluster -2.5 Add Data Storage to Data Center in step 2.1 (NFS storage) -2.6 Add ISO storage (NFS storage) -2.7 Create virtual machine (Create disk, disk attached), set network as dhcp mode, VM can get dhcp IP 3. Install rhvh-3.6-0.20170307.0 on another host (host 2) 4. Add host 2 (rhvh-3.6-0.20170307.0) to RHEVM 3.6 with the same cluster35(created in step 2.3) 5. Enable InClusterUpgrade option in RHEVM 3.6: # engine-config -s CheckMixedRhelVersions=false --cver=3.5 # service ovirt-engine restart In RHEVM 3.6 UI: -> Click the Clusters tab. -> Select the cluster35(created in step 2.3) and click Edit. -> Click the Scheduling Policy tab. -> Select "InClusterUpgrade" from the Select Policy drop-down list. 6. Migrate the virtual machine from host 1 to host 2 in RHEVM 3.6 UI: -> In the RHEVM 3.6 Virtual Machine tab -> Select the VM(created in step 2.7), click "Migrate" -> Select Destination host -> Select host 2 -> click "OK" Test Results: Migrate VM from el6-3.5-vintage to el7-3.6-ngn successfully in RHEVM 3.6 UI, and VM can get DHCP ip after migration.