Re: [Users] [node-devel] non-operational state as host does not meet clusters' minimu CPU level.
On Fri, Sep 21, 2012 at 7:01 AM, wujieke wuji...@qiyi.com wrote: [root@localhost ~]# virsh capabilities Please enter your authentication name: vdsm protects r/w libvirt socket, to avoid administrators accidentaly mess up w/ VMs under its control. You can use r/o connection virsh -r capabilities Cheers, Alan ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
[Users] Ceph / rbd and ovirt
I'm currently setting up an ovirt cluster and so far it looks good. I like the integration with Foreman http://theforeman.org/ . I would like to use Ceph / rbd for my storage. I saw some mention of patches coming in May, but I did not find any new posts. What is the status of this work? Is there some patches I can try out? I have a working Ceph cluster and a working ovirt cluster, I just need a way to bring them together. Thanks, JOSH ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [Users] Is there a way to force remove a host?
On 09/22/2012 11:19 PM, Eli Mesika wrote: - Original Message - From: Douglas Landgraf dougsl...@redhat.com To: Dominic Kaiser domi...@bostonvineyard.org Cc: Eli Mesika emes...@redhat.com, users@ovirt.org, Robert Middleswarth rob...@middleswarth.net Sent: Friday, September 21, 2012 8:12:27 PM Subject: Re: [Users] Is there a way to force remove a host? Hi Dominic, On 09/20/2012 12:11 PM, Dominic Kaiser wrote: Sorry I did not explain. I had tried to remove the host and had not luck troubleshooting it. I then had removed it and used it for a storage unit reinstalling fedora 17. I foolishly thought that I could just remove the host manually. It physically is not there. (My fault I know) Is there a way that you know of to remove a host brute force. dk Fell free to try the below script (not part of official project) for brute force: (from the engine side) # yum install python-psycopg2 -y # wget https://raw.github.com/dougsland/misc-rhev/master/engine_force_remove_Host.py # (edit the file and change the db password) # python ./engine_force_remove_Host.py Hi , had looked in the Python script you had provided: First, I must say that handling the database directly may leave DB in inconsistent state, therefore, if there is no other option, the database should be backed up prior to this operation. In addition, I do not like the execution of the SQL statements in the script. There is a SP called DeleteVds(v_vds_id UUID) and you should use that since it encapsulates all details. For example, your script does not handle permission clean-up as the SP does and therefore leaves garbage in the database. shouldn't foreign keys protect against this? (if not cascade deletes which i remember caused some issues with locks) In addition, a failure in your script may leave database in inconsistent state while the SP is executed in one transaction and will leave DB consistent. So, in short I would prefer in this case that the relevant SP will do the clean-up since this is the one that is used by the code and that insures (at least I hope so) , that all related entities are removed as well. Thanks -- Cheers Douglas ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [Users] Ceph / rbd and ovirt
On 09/22/2012 08:58 AM, Josh Logan wrote: I'm currently setting up an ovirt cluster and so far it looks good. I like the integration with Foreman http://theforeman.org/ . I would like to use Ceph / rbd for my storage. I saw some mention of patches coming in May, but I did not find any new posts. What is the status of this work? Is there some patches I can try out? I have a working Ceph cluster and a working ovirt cluster, I just need a way to bring them together. Thanks, JOSH I don't remember any active work on this right now (for sure nothing like the gluster integration being done). but iiuc, ceph provides posixfs support - did you try creating a posixfs based storage domain? (you would need a full host (not ovirt-node) to install ceph client components on). Thanks, Itamar ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [Users] non-operational state as host does not meet clusters' minimu CPU level.
On 09/21/2012 03:37 PM, wujieke wrote: currently, I create a cluster with Conroe family. My host/node works. And I can create some VMs on it. But I am not sure if the issue talking here will impact the VM's performance? Or stability ? performance only, and mostly for guests utilizing new westmere features. i think this patch was around this not sure if its in 4.10.0-7.fc17 or a new issue commit 498d1dd6f01a7f5488792c41f73f7cca57b2c890 Author: Gal Hammer gham...@redhat.com Date: Mon Jun 4 17:55:35 2012 +0300 BZ#737104 Disallow cross-vendor virtualization. Find vendor for all cpu models, including those based on another cpu module. Change-Id: Id52f757971e28058257d819036e58e0f1925c44d Signed-off-by: Gal Hammer gham...@redhat.com Reviewed-on: http://gerrit.ovirt.org/5035 Reviewed-by: Dan Kenigsberg dan...@redhat.com diff --git a/vdsm/caps.py b/vdsm/caps.py index d89099a..7aeedf8 100644 --- a/vdsm/caps.py +++ b/vdsm/caps.py @@ -103,20 +103,35 @@ def _getEmulatedMachines(): return [ m.firstChild.toxml() for m in guestTag.getElementsByTagName('machine') ] +def _getAllCpuModels(): +cpu_map = minidom.parseString( +file('/usr/share/libvirt/cpu_map.xml').read()) + +allModels = dict() +for m in cpu_map.getElementsByTagName('arch')[0].childNodes: +if m.nodeName != 'model': +continue +element = m.getElementsByTagName('vendor') +if element: +vendor = element[0].getAttribute('name') +else: +# If current model doesn't have a vendor, check if it has a model +# that it is based on. The models in the cpu_map.xml file are +# sorted in a way that the base model is always defined before. +element = m.getElementsByTagName('model') +if element: +vendor = allModels.get(element[0].getAttribute('name'), None) +else: +vendor = None +allModels[m.getAttribute('name')] = vendor + +return allModels + @utils.memoized def _getCompatibleCpuModels(): c = libvirtconnection.get() -cpu_map = minidom.parseString( -file('/usr/share/libvirt/cpu_map.xml').read()) -def vendor(modelElem): -vs = modelElem.getElementsByTagName('vendor') -if vs: -return vs[0].getAttribute('name') -else: -return None -allModels = [ (m.getAttribute('name'), vendor(m)) for m - in cpu_map.getElementsByTagName('arch')[0].childNodes - if m.nodeName == 'model' ] +allModels = _getAllCpuModels() + def compatible(model, vendor): if not vendor: return False @@ -133,7 +148,7 @@ def _getCompatibleCpuModels(): raise return [ 'model_' + model for (model, vendor) - in allModels if compatible(model, vendor) ] +in allModels.iteritems() if compatible(model, vendor) ] def _parseKeyVal(lines, delim='='): d = {} -Original Message- From: Itamar Heim [mailto:ih...@redhat.com] Sent: Friday, September 21, 2012 3:57 PM To: wujieke Cc: 'Mark Wu'; users@ovirt.org Subject: Re: [Users] non-operational state as host does not meet clusters' minimu CPU level. On 09/21/2012 09:24 AM, wujieke wrote: Thanks a lot. Mark. Attach output for reference. ok, so libvirt/virsh detect sandybridge, but vdsm only reports conroe. there used to be a bug around this in vdsm - which version of vdsm are you running? -Original Message- From: Mark Wu [mailto:wu...@linux.vnet.ibm.com] Sent: Friday, September 21, 2012 2:15 PM To: wujieke Cc: 'Itamar Heim'; users@ovirt.org Subject: Re: [Users] non-operational state as host does not meet clusters' minimu CPU level. On 09/21/2012 01:01 PM, wujieke wrote: I follow the wiki page to re-install ovirt with all-in-one version . my local host in ovirt is working now. Thanks a lot. Btw: the cmd virsh capabilities complains out : [root@localhost ~]# virsh capabilities Please enter your authentication name: Please enter your password: error: Failed to reconnect to the hypervisor error: no valid connection error: authentication failed: Failed to step SASL negotiation: -1 (SASL(-1): generic failure: All-whitespace username.) any idea? Please try virsh -r capabilities -Original Message- From: Itamar Heim [mailto:ih...@redhat.com] Sent: Friday, September 21, 2012 12:44 PM To: wujieke Cc: node-de...@ovirt.org; users@ovirt.org Subject: Re: [Users] non-operational state as host does not meet clusters' minimu CPU level. On 09/21/2012 03:54 AM, wujieke wrote: [root@localhost ~]# vdsClient -s 0 getVdsCaps | grep -i flags cpuFlags = fpu,vme,de,pse,tsc,msr,pae,mce,cx8,apic,sep,mtrr,pge,mca,cmov,pat,ps e 3 6,clfl ush,dts,acpi,mmx,fxsr,sse,sse2,ss,ht,tm,pbe,syscall,nx,pdpe1gb,rdtsc p , lm,con stant_tsc,arch_perfmon,pebs,bts,rep_good,nopl,xtopology,nonstop_tsc, a p erfmpe
Re: [Users] Can oVirt be installed in a virtual machine?
On 09/22/2012 05:35 PM, Nicolas Chenier wrote: *Question 1 - if oVirt goes down... do the ovirt-nodes and VMs remain up?* Can someone answer this please? :-) If the oVirt manager (ie. the web application running inside AS7) loses connectivity to the node, the VM's on that node will keep running. You should know; however, that the general design is for the manager to remain in contact with the nodes. Due to budget and space constraints, I currently have 2 servers total. What if I did the following: Server 1) Fedora 17 with KVM/Virt-manager... running oVirt as a VM (through virt-manager) off the iSCSI NAS. Fine Server 2) oVirt-node machine - one and only host machine for oVirt running on Server 1). Again fine. With this setup I can run VMs from iSCSI on oVirt-node Server 2). Yes, nearly identical to my setup. In the event that oVirt-node Server 2) goes down... is anything stopping me from setting up my VMs on Server 1) with the iSCSI storage from the NAS and run my VMs without oVirt through virt-manager? Yes, I don't think that will work out of the box. It could probably be done but it would require some manual steps. This would give me some form of redundancy (requiring manual intervention) in the event that my ovirt-node went down... is this a feasible setup? See previous comment. To make it even more redundant, maybe I should do the following with Server 2) Install Fedora 17 with KVM/Virt-Manager, and VDSM... in the event that Server 1) fails... I can run my VMs on Server 2) through virt-manager? Should I just drop oVirt for now and run virt-manager on my 2 hosts, moving VMs manually (as they are running off iSCSI NAS) if a host fails? tear It depends on what you are trying to do. oVirt and virt-manager solve different problems. I would say that virt-manager is probably OK for a small setup, but I wouldn't deploy an enterprise solution around it. You have enough gear for a small oVirt setup. Run with that and add more nodes as you can. My 2c. Thank you, Nic On Sat, Sep 22, 2012 at 4:09 PM, Keith Robertson krobe...@redhat.com mailto:krobe...@redhat.com wrote: On 09/22/2012 02:28 PM, Nicolas Chenier wrote: Question - if oVirt goes down... do the ovirt-nodes and VMs remain up? Keith, how would you set yourself up with these specs: 2 host servers (quad-core xeons with 32gigs of ram) Are you saying that you only have 2 machines in total, or that you have 2 machines that can be dedicated hypervisors (ie. ovirt-node) and a third machine that can be a dedicated manager? If the former then one machine must run some version of *nix compatible with oVirt Manager and, the other machine in this scenario can simply run ovirt-node. If the latter, then you have 1 box dedicated as a manager and 2 boxes as dedicated hypervisors. This is a fairly basic/good setup. 1 iSCSI NAS Starting to think there is no way to achieve HA with this setup? Not with only 2 boxes. No. oVirt requires a dedicated machine? Generally, speaking. Yes. Truly HA setups aren't cheap and people often have different ideas of what constitutes HA. Offhand I would think that you would need... - 2 boxes for the oVirt manager - Clustering software for the manager to facilitate an active/passive setup. - UPSs (at *least* 2) which can be controlled by clustering software. Why? Most clustering SW require a fence device. These will be your fence devices. - 2 boxes for your hypervisors (ie. ovirt-nodes). This will facilitate fail-over from one node to the other. HA isn't cheap and can't usually be done on 2 boxes, IMO unless you're failing over a single app. Thank you! Nic PS. Could oVirt be integrated into ovirt-node on every server? On Sat, Sep 22, 2012 at 2:16 PM, Keith Robertson krobe...@redhat.com mailto:krobe...@redhat.com wrote: On 09/22/2012 01:09 PM, Nicolas Chenier wrote: Hi Alan, I have oVirt running in a VM off my Desktop (Fedora 17 w/ KVM Virt-Manager) off my iSCSI NAS. I've attached Server #1 as my first host (it's running ovirt-node). In the process of setting up my storage domains. I have a few questions to the experts out there: 1) How do I add my CD .ISOs to setup new VMs? Create iSCSI storage domain? But then how do I copy my ISOs to it? Create an ISO storage domain and use the ovirt-iso-uploader to add your ISOs and .vfd files into that domain. 2) Can I run my oVirt VM from ovirt-node machine, without running it in oVirt (ie. setup iSCSI in virt-manager (as it is now) and run oVirt from virt-manager... then I can manage my hosts through that ovirt VM? Huh? You could run the oVirt Manager from a VM managed by virt-manager... yes. Running the oVirt manager inside
Re: [Users] vdsmd doesn't restart after rebooting
On Sun, Sep 23, 2012 at 03:42:38PM +0200, Joop wrote: Dan Kenigsberg wrote: I have the same problem, see for answers below. On Fri, Sep 21, 2012 at 11:36:01AM -0400, Douglas Landgraf wrote: Hi Nathanaƫl, On 09/21/2012 10:18 AM, Nathanaƫl Blanchet wrote: Hi all, In the latest vdsm build from git (vdsm-4.10.0-0.452.git87594e3.fc17.x86_64), vdsmd.service never starts alone after rebooting. Does it start well when you restart it later, manually? Do you see failed to connect to libvirt slightly up the log? manually restarting it works OK, service vdsmd restart I have had a look to journalctl anfd I've found this : systemd-vdsmd[538]: vdsm: Failed to define network filters on libvirt[FAILED] [root@node ~]# service vdsmd status Redirecting to /bin/systemctl status vdsmd.service vdsmd.service - Virtual Desktop Server Manager Loaded: loaded (/usr/lib/systemd/system/vdsmd.service; enabled) Active: failed (Result: exit-code) since Fri, 21 Sep 2012 12:13:01 +0200; 4min 56s ago Process: 543 ExecStart=/lib/systemd/systemd-vdsmd start (code=exited, status=1/FAILURE) CGroup: name=systemd:/system/vdsmd.service Sep 21 12:12:55 node.abes.fr systemd-vdsmd[543]: Note: Forwarding request to 'systemctl disable libvirt-guests.service'. Sep 21 12:12:56 node.abes.fr systemd-vdsmd[543]: vdsm: libvirt already configured for vdsm [ OK ] Sep 21 12:12:56 node.abes.fr systemd-vdsmd[543]: Starting wdmd... Sep 21 12:12:56 node.abes.fr systemd-vdsmd[543]: Starting sanlock... Sep 21 12:12:56 node.abes.fr systemd-vdsmd[543]: Starting iscsid: Sep 21 12:13:01 node.abes.fr systemd-vdsmd[543]: Starting libvirtd (via systemctl): [ OK ] May this nwfilter be the cause of the failure? If yes, do I need to open a BZ? Thanks for your report. If you can, please open a BZ. ... but please first check whether running python /usr/share/vdsm/nwfilter.pyc reports anything fishy. I don't have that file in /usr/share/vdsm rpm -aq | grep vdsm reports: vdsm-4.10.0-7.fc17.x86_64 vdsm-gluster-4.10.0-7.fc17.noarch vdsm-python-4.10.0-7.fc17.x86_64 vdsm-cli-4.10.0-7.fc17.noarch vdsm-xmlrpc-4.10.0-7.fc17.noarch vdsm-rest-4.10.0-7.fc17.noarch You are using an older code base, relative to the original poster, and it seems that the mode of failure is different. I suppose you do not have systemd-vdsmd[538]: vdsm: Failed to define network filters on libvirt[FAILED] but something else causing vdsm to fail to start. Could you look in the logs for clues? Regards, Dan. ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users