[ovirt-users] I want to switch back to local storage
Hello, At last night I have sussessfully migrated my vms from kvm to gluster based ovirt 3.5. It sounds good, but the results was terrible: The glusterfs (two node replicated) is too slow for our vms. It is a very big problem, so I have looking for soultion. I don't know why but glusterfsd now uses 500-800% of my cpus on each server. At this moment I have only one gluster volume for vms. I just wondering I will configure local storage on the second host and copy to all of vms from glusterfs to this volume. I know in this case the second host will be on a different Datacenter. But in this case what will happend with my glusterfs? It will be break it? Will I lost my data? Could it make split-brains on gluster? It will break the whole ovirt-portal? If possible I don't want to more downtime by this procedure. It is possible? My config: 2 node dell r710, raid5 , 3xbonded 1GBs NICs + one for internet access for vms. The glusterfs uses same network with ovirtmgmt. What can I do ? Thanks in advance Tibor ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] Migrating from Fedora 19 to Fedora 20 with oVirt 3.5.0
On Thu, Nov 6, 2014 at 7:18 PM, Gianluca Cecchi wrote: > > >> I've not tested all-in-one upgrade, just to be sure, save current >> iptables config and do a backup before starting the upgrade. >> >> >> >> > >> > Gianluca >> > >> >> >> -- >> Sandro Bonazzola >> Better technology. Faster innovation. Powered by community collaboration. >> See how it works at redhat.com >> > > > trying to proceed with migration to f20. > > [snip] > But I get this with "yum distro-sync" > > what does it mean? > > [root@tekkaman sysconfig]# yum distro-sync > > [snip] > --> Running transaction check > ---> Package SLOF.noarch 0:0.1.git20121018-1.fc19 will be a downgrade > ---> Package SLOF.noarch 0:0.1.git20130430-2.fc19 will be erased > ---> Package ipxe-roms-qemu.noarch 0:20130517-2.gitc4bce43.fc19 will be a > downgrade > ---> Package ipxe-roms-qemu.noarch 0:20130517-3.gitc4bce43.fc19 will be > erased > ---> Package libcacard.x86_64 2:1.4.2-15.fc19 will be a downgrade > ---> Package libcacard.x86_64 2:1.6.1-2.fc19 will be erased > ---> Package openbios.noarch 0:1.0.svn1063-2.fc19 will be a downgrade > ---> Package openbios.noarch 0:1.1.svn1198-2.fc19 will be erased > ---> Package qemu.x86_64 2:1.4.2-15.fc19 will be a downgrade > ---> Package qemu.x86_64 2:1.6.1-2.fc19 will be erased > ---> Package qemu-common.x86_64 2:1.4.2-15.fc19 will be a downgrade > ---> Package qemu-common.x86_64 2:1.6.1-2.fc19 will be erased > ---> Package qemu-img.x86_64 2:1.4.2-15.fc19 will be a downgrade > ---> Package qemu-img.x86_64 2:1.6.1-2.fc19 will be erased > ---> Package qemu-kvm.x86_64 2:1.4.2-15.fc19 will be a downgrade > ---> Package qemu-kvm.x86_64 2:1.6.1-2.fc19 will be erased > In practice the problem is due to the fact that it is an all-in-one installation and so it pulled in the virt-preview part for vdsm and related packages (qemu-kvm, seabioes, etc..). How much is it risky to bypass the "yum distro-sync" step? Will the fedup part run itself the distro-sync command then, or is it only a sort of pre-check to verify consistency of packages? In my case if fedup will retain the ovirt repos all should go ok, correct? Gianluca ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] how to 'reset' a failed install?
On 11/08/2014 04:55 AM, Yair Zaslavsky wrote: > > - Original Message - >> From: "Robert Story" >> To: users@ovirt.org >> Sent: Saturday, November 8, 2014 3:09:02 AM >> Subject: [ovirt-users] how to 'reset' a failed install? >> >> I've been doing lots of unsuccessful 3.5 hosted-engine installs in my lab, >> where it's easy for me to re-install the OS if I need to start over. Now I >> need to try an install in a remote datacenter where I won't be able to >> re-install the OS. So I was wondering if there is a way to 'reset' a failed >> install so that another install can be attempted... >> >> My thoughts so far are: >> >> - stop vdsm, supervdsm, and libvirt >> - use etckeeper to reset everything under /etc >> - delete old log files >> - delete hosted_engine storage domain on storage (if install got that far) >> - restart vdsm, supervdsm, and libvirt >> >> What am I missing? Maybe some remnants in /var (hmm, probably the vdsm >> persistent config)? Anything else? Why not use engine-cleanup and be done with it? Or doesn't it work on hosted engine? Kind regards, Jorick Met vriendelijke groet, With kind regards, Jorick Astrego Netbulae Virtualization Experts Tel: 053 20 30 270 i...@netbulae.euStaalsteden 4-3A KvK 08198180 Fax: 053 20 30 271 www.netbulae.eu 7547 TA Enschede BTW NL821234584B01 ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] webhook
- Original Message -
> From: "Vojtech Szocs"
> To: "Yair Zaslavsky"
> Cc: "Barak Azulay" , "Oved Ourfali" ,
> users@ovirt.org
> Sent: Friday, November 7, 2014 5:16:48 PM
> Subject: Re: [ovirt-users] webhook
>
>
>
> - Original Message -
> > From: "Yair Zaslavsky"
> > To: "Vojtech Szocs"
> > Cc: "Barak Azulay" , "Oved Ourfali" ,
> > users@ovirt.org
> > Sent: Thursday, November 6, 2014 2:59:53 PM
> > Subject: Re: [ovirt-users] webhook
> >
> >
> >
> > - Original Message -
> > > From: "Vojtech Szocs"
> > > To: "Barak Azulay"
> > > Cc: "Oved Ourfali" , users@ovirt.org
> > > Sent: Thursday, November 6, 2014 3:38:56 PM
> > > Subject: Re: [ovirt-users] webhook
> > >
> > >
> > >
> > > - Original Message -
> > > > From: "Barak Azulay"
> > > > To: "Vojtech Szocs"
> > > > Cc: "Einav Cohen" , "Oved Ourfali"
> > > > ,
> > > > users@ovirt.org
> > > > Sent: Tuesday, November 4, 2014 5:15:35 PM
> > > > Subject: Re: [ovirt-users] webhook
> > > >
> > > >
> > > >
> > > > - Original Message -
> > > > > From: "Vojtech Szocs"
> > > > > To: "Einav Cohen"
> > > > > Cc: "Oved Ourfali" , users@ovirt.org
> > > > > Sent: Tuesday, November 4, 2014 2:12:05 PM
> > > > > Subject: Re: [ovirt-users] webhook
> > > > >
> > > > >
> > > > >
> > > > > - Original Message -
> > > > > > From: "Einav Cohen"
> > > > > > To: "Vojtech Szocs"
> > > > > > Cc: "Oved Ourfali" , users@ovirt.org
> > > > > > Sent: Friday, October 31, 2014 8:01:34 PM
> > > > > > Subject: Re: [ovirt-users] webhook
> > > > > >
> > > > > > > - Original Message -
> > > > > > > From: "Vojtech Szocs"
> > > > > > > Sent: Friday, October 31, 2014 11:51:53 AM
> > > > > > >
> > > > > > > Hi,
> > > > > > >
> > > > > > > if I get this correctly, you'd like to be notified when certain
> > > > > > > event
> > > > > > > happens (VM created/deleted/etc.) and react upon that. I see
> > > > > > > multiple
> > > > > > > possible approaches here:
> > > > > > >
> > > > > > > 0, improve Engine extension API (refer to Alon Bar-Lev for
> > > > > > > details)
> > > > > > >- if extensions can be packaged as JARs and these JARs could
> > > > > > >include
> > > > > > > web fragments [1] it would mean the possibility to deploy
> > > > > > > custom
> > > > > > > servlets onto existing Engine instance (in context of webapp
> > > > > > > that
> > > > > > > processes extensions)
> > > > > > >- your custom Java servlet could query REST interface (or be
> > > > > > >notified
> > > > > > > once something happens, but AFAIK we don't have that
> > > > > > > implemented
> > > > > > > yet)
> > > > > > > and do whatever logic is needed
> > > > > > >- once I asked Alon about ^^ but never got response from him
> > > > > > >- IMHO this would be a nice way to deploy custom Java code on
> > > > > > >Engine
> > > > > > >
> > > > > > > [1]
> > > > > > > https://blogs.oracle.com/swchan/entry/servlet_3_0_web_fragment
> > > > > > >
> > > > > > > 1, improve UI plugin API
> > > > > > >- add "VirtualMachineDataLoaded" event fired upon each refresh
> > > > > > >of
> > > > > > > VM data in UI table (generalization -> "{Entity}DataLoaded")
> > > > > > >- this is similar to existing "{Entity}SelectionChange" events
> > > > > >
> > > > > > relying on changes in the UI table is a bad idea:
> > > > > >
> > > > > > (1) potentially missing events:
> > > > > > the UI displays paginated data; if my VMs are sorted by name, and
> > > > > > I have 1000 VMs in my setup, and I just added a VM named "z", it
> > > > > > will
> > > > > > be added to the last "page" which is not displayed right now, so I
> > > > > > wouldn't even be aware that something was added.
> > > > > >
> > > > > > (2) potentially "creating" "fake" events:
> > > > > > changes in the displayed data in the UI can occur due to change in
> > > > > > the
> > > > > > Search query; if I have 50 VMs in my setup, and I initially had the
> > > > > > "Vms:" search query, and now I change it to "VMs: name = a*", which
> > > > > > results in displaying only 10 VMs, this may falsely hint on removal
> > > > > > of 40 VMs from the system.
> > > > >
> > > > > Agreed, that was a bad idea :) UI data is constrained by things like
> > > > > search query, pagination, sorting etc. Scratch my previous idea.
> > > > >
> > > > > As mentioned before, notification of relevant events occuring in
> > > > > system
> > > > > should be the way to go. Either Engine will provide mechanism to
> > > > > notify
> > > > > other systems (i.e. email notifier) or other system will
> > > > > poll/listen-to
> > > > > changes in Engine (i.e. via REST API).
> > > > >
> > > > > Idea for consideration, if every change to oVirt system would be sent
> > > > > to some event bus, we could easily implement different notification
> > > > > mechanisms (like websocket in addition to email), and web apps living
> > > > > in Engine EAR context could
