Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
Thanks for the clarification, is there a bug tracking this in libvirt already? Actually I don't think there is one, so feel free to file one I took the liberty of doing so: https://bugzilla.redhat.com/show_bug.cgi?id=1208588 On Wed, Mar 18, 2015 at 6:11 PM, Daniel P. Berrange berra...@redhat.com wrote: On Wed, Mar 18, 2015 at 10:59:19AM -0700, Joe Gordon wrote: On Wed, Mar 18, 2015 at 3:09 AM, Daniel P. Berrange berra...@redhat.com wrote: On Wed, Mar 18, 2015 at 08:33:26AM +0100, Thomas Herve wrote: Interesting bug. I think I agree with you that there isn't a good solution currently for instances that have a mix of shared and not-shared storage. I'm curious what Daniel meant by saying that marking the disk shareable is not as reliable as we would want. I think this is the bug I reported here: https://bugs.launchpad.net/nova/+bug/1376615 My initial approach was indeed to mark the disks are shareable: the patch (https://review.openstack.org/#/c/125616/) has comments around the issues, mainly around I/Ocache and SELinux isolation being disabled. Yep, those are both show stopper issues. The only solution is to fix the libvirt API for this first. Thanks for the clarification, is there a bug tracking this in libvirt already? Actually I don't think there is one, so feel free to file one Regards, Daniel -- |: http://berrange.com -o-http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :| __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
On Wed, Mar 18, 2015 at 08:33:26AM +0100, Thomas Herve wrote: Interesting bug. I think I agree with you that there isn't a good solution currently for instances that have a mix of shared and not-shared storage. I'm curious what Daniel meant by saying that marking the disk shareable is not as reliable as we would want. I think this is the bug I reported here: https://bugs.launchpad.net/nova/+bug/1376615 My initial approach was indeed to mark the disks are shareable: the patch (https://review.openstack.org/#/c/125616/) has comments around the issues, mainly around I/Ocache and SELinux isolation being disabled. Yep, those are both show stopper issues. The only solution is to fix the libvirt API for this first. Regards, Daniel -- |: http://berrange.com -o-http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :| __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
On Tue, Mar 17, 2015 at 01:33:26PM -0700, Joe Gordon wrote: On Thu, Jun 19, 2014 at 1:38 AM, Daniel P. Berrange berra...@redhat.com wrote: On Wed, Jun 18, 2014 at 11:09:33PM -0700, Rafi Khardalian wrote: I am concerned about how block migration functions when Cinder volumes are attached to an instance being migrated. We noticed some unexpected behavior recently, whereby attached generic NFS-based volumes would become entirely unsparse over the course of a migration. After spending some time reviewing the code paths in Nova, I'm more concerned that this was actually a minor symptom of a much more significant issue. For those unfamiliar, NFS-based volumes are simply RAW files residing on an NFS mount. From Libvirt's perspective, these volumes look no different than root or ephemeral disks. We are currently not filtering out volumes whatsoever when making the request into Libvirt to perform the migration. Libvirt simply receives an additional flag (VIR_MIGRATE_NON_SHARED_INC) when a block migration is requested, which applied to the entire migration process, not differentiated on a per-disk basis. Numerous guards within Nova to prevent a block based migration from being allowed if the instance disks exist on the destination; yet volumes remain attached and within the defined XML during a block migration. Unless Libvirt has a lot more logic around this than I am lead to believe, this seems like a recipe for corruption. It seems as though this would also impact any type of volume attached to an instance (iSCSI, RBD, etc.), NFS just happens to be what we were testing. If I am wrong and someone can correct my understanding, I would really appreciate it. Otherwise, I'm surprised we haven't had more reports of issues when block migrations are used in conjunction with any attached volumes. Libvirt/QEMU has no special logic. When told to block-migrate, it will do so for *all* disks attached to the VM in read-write-exclusive mode. It will only skip those marked read-only or read-write-shared mode. Even that distinction is somewhat dubious and so not reliably what you would want. It seems like we should just disallow block migrate when any cinder volumes are attached to the VM, since there is never any valid use case for doing block migrate from a cinder volume to itself. Digging up this old thread because I am working on getting multi node live migration testing working (https://review.openstack.org/#/c/165182/), and just ran into this issue (bug 1398999). And I am not sure I agree with this statement. I think there is a valid case for doing block migrate with a cinder volume attached to an instance: To be clear, I'm not saying the use cases for block migrating cinder are invalid. Just that with the way libvirt exposes block migration today it isn't safe for us to allow it, because we don't have fine grained control to make it reliably safe from openstack. We need to improve the libvirt API in this area and then we can support this feature properly. * Cloud isn't using a shared filesystem for ephemeral storage * Instance is booted from an image, and a volume is attached afterwards. An admin wants to take the box the instance is running on offline for maintanince with a minimal impact to the instances running on it. What is the recommended solution for that use case? If the admin disconnects and reconnects the volume themselves is there a risk of impacting whats running on the instance? etc. Yes, and that sucks, but that's the only safe option today, otherwise libvirt is going to try copying the data in the cinder volumes itself, which means it is copying from the volume on one host, back into the very same volume on the other host. IOW it is rewriting all the data even though the volume is shared betwteen the hosts. This has dangerous data corruption failure scenarios as well as being massively wasteful of CPU and network bandwidth. Regards, Daniel -- |: http://berrange.com -o-http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :| __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
On Wed, Mar 18, 2015 at 3:09 AM, Daniel P. Berrange berra...@redhat.com wrote: On Wed, Mar 18, 2015 at 08:33:26AM +0100, Thomas Herve wrote: Interesting bug. I think I agree with you that there isn't a good solution currently for instances that have a mix of shared and not-shared storage. I'm curious what Daniel meant by saying that marking the disk shareable is not as reliable as we would want. I think this is the bug I reported here: https://bugs.launchpad.net/nova/+bug/1376615 My initial approach was indeed to mark the disks are shareable: the patch (https://review.openstack.org/#/c/125616/) has comments around the issues, mainly around I/Ocache and SELinux isolation being disabled. Yep, those are both show stopper issues. The only solution is to fix the libvirt API for this first. Thanks for the clarification, is there a bug tracking this in libvirt already? Regards, Daniel -- |: http://berrange.com -o-http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :| __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
Interesting bug. I think I agree with you that there isn't a good solution currently for instances that have a mix of shared and not-shared storage. I'm curious what Daniel meant by saying that marking the disk shareable is not as reliable as we would want. I think this is the bug I reported here: https://bugs.launchpad.net/nova/+bug/1376615 My initial approach was indeed to mark the disks are shareable: the patch (https://review.openstack.org/#/c/125616/) has comments around the issues, mainly around I/Ocache and SELinux isolation being disabled. -- Thomas __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
On Wed, Mar 18, 2015 at 10:59:19AM -0700, Joe Gordon wrote: On Wed, Mar 18, 2015 at 3:09 AM, Daniel P. Berrange berra...@redhat.com wrote: On Wed, Mar 18, 2015 at 08:33:26AM +0100, Thomas Herve wrote: Interesting bug. I think I agree with you that there isn't a good solution currently for instances that have a mix of shared and not-shared storage. I'm curious what Daniel meant by saying that marking the disk shareable is not as reliable as we would want. I think this is the bug I reported here: https://bugs.launchpad.net/nova/+bug/1376615 My initial approach was indeed to mark the disks are shareable: the patch (https://review.openstack.org/#/c/125616/) has comments around the issues, mainly around I/Ocache and SELinux isolation being disabled. Yep, those are both show stopper issues. The only solution is to fix the libvirt API for this first. Thanks for the clarification, is there a bug tracking this in libvirt already? Actually I don't think there is one, so feel free to file one Regards, Daniel -- |: http://berrange.com -o-http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :| __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
On Thu, Jun 19, 2014 at 1:38 AM, Daniel P. Berrange berra...@redhat.com wrote: On Wed, Jun 18, 2014 at 11:09:33PM -0700, Rafi Khardalian wrote: I am concerned about how block migration functions when Cinder volumes are attached to an instance being migrated. We noticed some unexpected behavior recently, whereby attached generic NFS-based volumes would become entirely unsparse over the course of a migration. After spending some time reviewing the code paths in Nova, I'm more concerned that this was actually a minor symptom of a much more significant issue. For those unfamiliar, NFS-based volumes are simply RAW files residing on an NFS mount. From Libvirt's perspective, these volumes look no different than root or ephemeral disks. We are currently not filtering out volumes whatsoever when making the request into Libvirt to perform the migration. Libvirt simply receives an additional flag (VIR_MIGRATE_NON_SHARED_INC) when a block migration is requested, which applied to the entire migration process, not differentiated on a per-disk basis. Numerous guards within Nova to prevent a block based migration from being allowed if the instance disks exist on the destination; yet volumes remain attached and within the defined XML during a block migration. Unless Libvirt has a lot more logic around this than I am lead to believe, this seems like a recipe for corruption. It seems as though this would also impact any type of volume attached to an instance (iSCSI, RBD, etc.), NFS just happens to be what we were testing. If I am wrong and someone can correct my understanding, I would really appreciate it. Otherwise, I'm surprised we haven't had more reports of issues when block migrations are used in conjunction with any attached volumes. Libvirt/QEMU has no special logic. When told to block-migrate, it will do so for *all* disks attached to the VM in read-write-exclusive mode. It will only skip those marked read-only or read-write-shared mode. Even that distinction is somewhat dubious and so not reliably what you would want. It seems like we should just disallow block migrate when any cinder volumes are attached to the VM, since there is never any valid use case for doing block migrate from a cinder volume to itself. Digging up this old thread because I am working on getting multi node live migration testing working (https://review.openstack.org/#/c/165182/), and just ran into this issue (bug 1398999). And I am not sure I agree with this statement. I think there is a valid case for doing block migrate with a cinder volume attached to an instance: * Cloud isn't using a shared filesystem for ephemeral storage * Instance is booted from an image, and a volume is attached afterwards. An admin wants to take the box the instance is running on offline for maintanince with a minimal impact to the instances running on it. What is the recommended solution for that use case? If the admin disconnects and reconnects the volume themselves is there a risk of impacting whats running on the instance? etc. Regards, Daniel -- |: http://berrange.com -o-http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :| ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
On 03/17/2015 02:33 PM, Joe Gordon wrote: Digging up this old thread because I am working on getting multi node live migration testing working (https://review.openstack.org/#/c/165182/), and just ran into this issue (bug 1398999). And I am not sure I agree with this statement. I think there is a valid case for doing block migrate with a cinder volume attached to an instance: * Cloud isn't using a shared filesystem for ephemeral storage * Instance is booted from an image, and a volume is attached afterwards. An admin wants to take the box the instance is running on offline for maintanince with a minimal impact to the instances running on it. What is the recommended solution for that use case? If the admin disconnects and reconnects the volume themselves is there a risk of impacting whats running on the instance? etc. Interesting bug. I think I agree with you that there isn't a good solution currently for instances that have a mix of shared and not-shared storage. I'm curious what Daniel meant by saying that marking the disk shareable is not as reliable as we would want. I think there is definitely a risk if the admin disconnects the volume--whether or not that causes problems depends on whether the application can handle that cleanly. I suspect the proper cloud-aware strategy would be to just kill it and have another instance take over. But that's not very helpful for not-fully-cloud-aware applications. Also, since you've been playing in this area...do you know if we currently properly support all variations on live/cold migration, resize, evacuate, etc. for the boot-from-volume case? Chris __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
On Mon, Feb 16, 2015 at 01:39:21PM +1300, Robert Collins wrote: On 19 June 2014 at 20:38, Daniel P. Berrange berra...@redhat.com wrote: On Wed, Jun 18, 2014 at 11:09:33PM -0700, Rafi Khardalian wrote: I am concerned about how block migration functions when Cinder volumes are attached to an instance being migrated. We noticed some unexpected behavior recently, whereby attached generic NFS-based volumes would become entirely unsparse over the course of a migration. After spending some time reviewing the code paths in Nova, I'm more concerned that this was actually a minor symptom of a much more significant issue. For those unfamiliar, NFS-based volumes are simply RAW files residing on an NFS mount. From Libvirt's perspective, these volumes look no different than root or ephemeral disks. We are currently not filtering out volumes whatsoever when making the request into Libvirt to perform the migration. Libvirt simply receives an additional flag (VIR_MIGRATE_NON_SHARED_INC) when a block migration is requested, which applied to the entire migration process, not differentiated on a per-disk basis. Numerous guards within Nova to prevent a block based migration from being allowed if the instance disks exist on the destination; yet volumes remain attached and within the defined XML during a block migration. Unless Libvirt has a lot more logic around this than I am lead to believe, this seems like a recipe for corruption. It seems as though this would also impact any type of volume attached to an instance (iSCSI, RBD, etc.), NFS just happens to be what we were testing. If I am wrong and someone can correct my understanding, I would really appreciate it. Otherwise, I'm surprised we haven't had more reports of issues when block migrations are used in conjunction with any attached volumes. Libvirt/QEMU has no special logic. When told to block-migrate, it will do so for *all* disks attached to the VM in read-write-exclusive mode. It will only skip those marked read-only or read-write-shared mode. Even that distinction is somewhat dubious and so not reliably what you would want. It seems like we should just disallow block migrate when any cinder volumes are attached to the VM, since there is never any valid use case for doing block migrate from a cinder volume to itself. Regards, Daniel Just ran across this from bug https://bugs.launchpad.net/nova/+bug/1398999. Is there some way to signal to libvirt that some block devices shouldn't be migrated by it but instead are known to be networked etc? Or put another way, how can we have our cake and eat it too. Its not uncommon for a VM to be cinder booted but have local storage for swap... and AIUI the fix we put in for this bug stops those VM's being migrated. Do you think it is tractable (but needs libvirt work), or is it something endemic to the problem (e.g. dirty page synchronisation with the VM itself) that will be in the way? It is merely a missing feature in libvirt that no one has had the time to address yet. Regards, Daniel -- |: http://berrange.com -o-http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :| __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
On 19 June 2014 at 20:38, Daniel P. Berrange berra...@redhat.com wrote: On Wed, Jun 18, 2014 at 11:09:33PM -0700, Rafi Khardalian wrote: I am concerned about how block migration functions when Cinder volumes are attached to an instance being migrated. We noticed some unexpected behavior recently, whereby attached generic NFS-based volumes would become entirely unsparse over the course of a migration. After spending some time reviewing the code paths in Nova, I'm more concerned that this was actually a minor symptom of a much more significant issue. For those unfamiliar, NFS-based volumes are simply RAW files residing on an NFS mount. From Libvirt's perspective, these volumes look no different than root or ephemeral disks. We are currently not filtering out volumes whatsoever when making the request into Libvirt to perform the migration. Libvirt simply receives an additional flag (VIR_MIGRATE_NON_SHARED_INC) when a block migration is requested, which applied to the entire migration process, not differentiated on a per-disk basis. Numerous guards within Nova to prevent a block based migration from being allowed if the instance disks exist on the destination; yet volumes remain attached and within the defined XML during a block migration. Unless Libvirt has a lot more logic around this than I am lead to believe, this seems like a recipe for corruption. It seems as though this would also impact any type of volume attached to an instance (iSCSI, RBD, etc.), NFS just happens to be what we were testing. If I am wrong and someone can correct my understanding, I would really appreciate it. Otherwise, I'm surprised we haven't had more reports of issues when block migrations are used in conjunction with any attached volumes. Libvirt/QEMU has no special logic. When told to block-migrate, it will do so for *all* disks attached to the VM in read-write-exclusive mode. It will only skip those marked read-only or read-write-shared mode. Even that distinction is somewhat dubious and so not reliably what you would want. It seems like we should just disallow block migrate when any cinder volumes are attached to the VM, since there is never any valid use case for doing block migrate from a cinder volume to itself. Regards, Daniel Just ran across this from bug https://bugs.launchpad.net/nova/+bug/1398999. Is there some way to signal to libvirt that some block devices shouldn't be migrated by it but instead are known to be networked etc? Or put another way, how can we have our cake and eat it too. Its not uncommon for a VM to be cinder booted but have local storage for swap... and AIUI the fix we put in for this bug stops those VM's being migrated. Do you think it is tractable (but needs libvirt work), or is it something endemic to the problem (e.g. dirty page synchronisation with the VM itself) that will be in the way? -Rob -- Robert Collins rbtcoll...@hp.com Distinguished Technologist HP Converged Cloud __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
On Mon, Feb 16, 2015 at 01:39:21PM +1300, Robert Collins wrote: Just ran across this from bug https://bugs.launchpad.net/nova/+bug/1398999. Is there some way to signal to libvirt that some block devices shouldn't be migrated by it but instead are known to be networked etc? Or put another way, how can we have our cake and eat it too. Its not uncommon for a VM to be cinder booted but have local storage for swap... and AIUI the fix we put in for this bug stops those VM's being migrated. Do you think it is tractable (but needs libvirt work), or is it something endemic to the problem (e.g. dirty page synchronisation with the VM itself) that will be in the way? I have a half drafted email for the libvirt devel list proposing this exact thing. Allow an element in the XML that tell libvirt how/if it can/should migrate a device. As noted previously I'm happy to do qemu/libvirt work to help openstack. Doign this would solve a few other issues in openstack without breaking existing users. So now I look like I'm jumping on your bandwagon Phooey Yours Tony. pgpryzA1nk4rz.pgp Description: PGP signature __ OpenStack Development Mailing List (not for usage questions) Unsubscribe: openstack-dev-requ...@lists.openstack.org?subject:unsubscribe http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
On Wed, Jun 18, 2014 at 11:09:33PM -0700, Rafi Khardalian wrote: I am concerned about how block migration functions when Cinder volumes are attached to an instance being migrated. We noticed some unexpected behavior recently, whereby attached generic NFS-based volumes would become entirely unsparse over the course of a migration. After spending some time reviewing the code paths in Nova, I'm more concerned that this was actually a minor symptom of a much more significant issue. For those unfamiliar, NFS-based volumes are simply RAW files residing on an NFS mount. From Libvirt's perspective, these volumes look no different than root or ephemeral disks. We are currently not filtering out volumes whatsoever when making the request into Libvirt to perform the migration. Libvirt simply receives an additional flag (VIR_MIGRATE_NON_SHARED_INC) when a block migration is requested, which applied to the entire migration process, not differentiated on a per-disk basis. Numerous guards within Nova to prevent a block based migration from being allowed if the instance disks exist on the destination; yet volumes remain attached and within the defined XML during a block migration. Unless Libvirt has a lot more logic around this than I am lead to believe, this seems like a recipe for corruption. It seems as though this would also impact any type of volume attached to an instance (iSCSI, RBD, etc.), NFS just happens to be what we were testing. If I am wrong and someone can correct my understanding, I would really appreciate it. Otherwise, I'm surprised we haven't had more reports of issues when block migrations are used in conjunction with any attached volumes. Libvirt/QEMU has no special logic. When told to block-migrate, it will do so for *all* disks attached to the VM in read-write-exclusive mode. It will only skip those marked read-only or read-write-shared mode. Even that distinction is somewhat dubious and so not reliably what you would want. It seems like we should just disallow block migrate when any cinder volumes are attached to the VM, since there is never any valid use case for doing block migrate from a cinder volume to itself. Regards, Daniel -- |: http://berrange.com -o-http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :| ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
The use-case for block migration in Libvirt/QEMU is to allow migration between two different back-ends. This is basically a host based volume migration, ESXi has a similar functionality (storage vMotion), but probably not enabled with OpenStack. Btw, if the Cinder volume driver can migrate the volume by itself, the Libvirt/QEMU is not called upon, but if it can't (different vendor boxes don't talk to each other), then Cinder asks Nova to help move the data... If you are missing this host based process you are basically have a data lock-in on a specific back-end - the use case could be storage evacuation, or just moving the data to a different box. Ronen, From: Daniel P. Berrange berra...@redhat.com To: OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, Date: 19/06/2014 11:42 AM Subject:Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes On Wed, Jun 18, 2014 at 11:09:33PM -0700, Rafi Khardalian wrote: I am concerned about how block migration functions when Cinder volumes are attached to an instance being migrated. We noticed some unexpected behavior recently, whereby attached generic NFS-based volumes would become entirely unsparse over the course of a migration. After spending some time reviewing the code paths in Nova, I'm more concerned that this was actually a minor symptom of a much more significant issue. For those unfamiliar, NFS-based volumes are simply RAW files residing on an NFS mount. From Libvirt's perspective, these volumes look no different than root or ephemeral disks. We are currently not filtering out volumes whatsoever when making the request into Libvirt to perform the migration. Libvirt simply receives an additional flag (VIR_MIGRATE_NON_SHARED_INC) when a block migration is requested, which applied to the entire migration process, not differentiated on a per-disk basis. Numerous guards within Nova to prevent a block based migration from being allowed if the instance disks exist on the destination; yet volumes remain attached and within the defined XML during a block migration. Unless Libvirt has a lot more logic around this than I am lead to believe, this seems like a recipe for corruption. It seems as though this would also impact any type of volume attached to an instance (iSCSI, RBD, etc.), NFS just happens to be what we were testing. If I am wrong and someone can correct my understanding, I would really appreciate it. Otherwise, I'm surprised we haven't had more reports of issues when block migrations are used in conjunction with any attached volumes. Libvirt/QEMU has no special logic. When told to block-migrate, it will do so for *all* disks attached to the VM in read-write-exclusive mode. It will only skip those marked read-only or read-write-shared mode. Even that distinction is somewhat dubious and so not reliably what you would want. It seems like we should just disallow block migrate when any cinder volumes are attached to the VM, since there is never any valid use case for doing block migrate from a cinder volume to itself. Regards, Daniel -- |: http://berrange.com -o-http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :| ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes
I think here there are two different processes making us of libvirt block migration: 1) Instance migration, which should not do anything with cinder volumes 2) Cinder live migration between backends, which is what I think Ronen Kat is referring to On 19 June 2014 11:06, Ronen Kat ronen...@il.ibm.com wrote: The use-case for block migration in Libvirt/QEMU is to allow migration between two different back-ends. This is basically a host based volume migration, ESXi has a similar functionality (storage vMotion), but probably not enabled with OpenStack. Btw, if the Cinder volume driver can migrate the volume by itself, the Libvirt/QEMU is not called upon, but if it can't (different vendor boxes don't talk to each other), then Cinder asks Nova to help move the data... If you are missing this host based process you are basically have a data lock-in on a specific back-end - the use case could be storage evacuation, or just moving the data to a different box. Ronen, From:Daniel P. Berrange berra...@redhat.com To:OpenStack Development Mailing List (not for usage questions) openstack-dev@lists.openstack.org, Date:19/06/2014 11:42 AM Subject:Re: [openstack-dev] [nova][libvirt] Block migrations and Cinder volumes On Wed, Jun 18, 2014 at 11:09:33PM -0700, Rafi Khardalian wrote: I am concerned about how block migration functions when Cinder volumes are attached to an instance being migrated. We noticed some unexpected behavior recently, whereby attached generic NFS-based volumes would become entirely unsparse over the course of a migration. After spending some time reviewing the code paths in Nova, I'm more concerned that this was actually a minor symptom of a much more significant issue. For those unfamiliar, NFS-based volumes are simply RAW files residing on an NFS mount. From Libvirt's perspective, these volumes look no different than root or ephemeral disks. We are currently not filtering out volumes whatsoever when making the request into Libvirt to perform the migration. Libvirt simply receives an additional flag (VIR_MIGRATE_NON_SHARED_INC) when a block migration is requested, which applied to the entire migration process, not differentiated on a per-disk basis. Numerous guards within Nova to prevent a block based migration from being allowed if the instance disks exist on the destination; yet volumes remain attached and within the defined XML during a block migration. Unless Libvirt has a lot more logic around this than I am lead to believe, this seems like a recipe for corruption. It seems as though this would also impact any type of volume attached to an instance (iSCSI, RBD, etc.), NFS just happens to be what we were testing. If I am wrong and someone can correct my understanding, I would really appreciate it. Otherwise, I'm surprised we haven't had more reports of issues when block migrations are used in conjunction with any attached volumes. Libvirt/QEMU has no special logic. When told to block-migrate, it will do so for *all* disks attached to the VM in read-write-exclusive mode. It will only skip those marked read-only or read-write-shared mode. Even that distinction is somewhat dubious and so not reliably what you would want. It seems like we should just disallow block migrate when any cinder volumes are attached to the VM, since there is never any valid use case for doing block migrate from a cinder volume to itself. Regards, Daniel -- |: http://berrange.com -o-http://www.flickr.com/photos/dberrange/ :| |: http://libvirt.org -o- http://virt-manager.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :| ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev -- Duncan Thomas ___ OpenStack-dev mailing list OpenStack-dev@lists.openstack.org http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
