From mboxrd@z Thu Jan 1 00:00:00 1970 From: Hongyang Yang Subject: Re: [PATCH Remus v2 00/10] Remus support for Migration-v2 Date: Mon, 11 May 2015 14:28:48 +0800 Message-ID: <55504C20.8020009@cn.fujitsu.com> References: <1431077610-3366-1-git-send-email-yanghy@cn.fujitsu.com> <554CFC90.1030301@citrix.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; Format="flowed" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <554CFC90.1030301@citrix.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Andrew Cooper , xen-devel@lists.xen.org Cc: wei.liu2@citrix.com, ian.campbell@citrix.com, wency@cn.fujitsu.com, ian.jackson@eu.citrix.com, yunhong.jiang@intel.com, eddie.dong@intel.com, rshriram@cs.ubc.ca List-Id: xen-devel@lists.xenproject.org On 05/09/2015 02:12 AM, Andrew Cooper wrote: > On 08/05/15 10:33, Yang Hongyang wrote: >> This patchset implement the Remus support for Migration v2 but without >> memory compressing. >> >> The series can be found on github: >> https://github.com/macrosheep/xen/tree/Remus-newmig-v2 >> >> PATCH 1-7: Some refactor and prepare work. >> PATCH 8-9: The main Remus loop implement. >> PATCH 10: Fix for Remus. > > I have reviewed the other half of the series now, and have some design > to discuss. (I was hoping to get this email sent in reply to v1, but > never mind). This largely concerns patch 7 and onwards. > > Migration v2 has substantially more structure than legacy did. Once > issue so far is that your series relies on using more than one END > record, which is not supported in the spec. (Of course - the spec is > fine to be extended in forward-compatible ways.) I use END record as a info that indicate the end of the stream. I saw that you add a checkpoint record in your v2 series of Remus related patches, I can use that record to indicate the end of the checkpointed stream, but I think the record better to be called as end-of-checkpoint? > > To fix the qemu layering issues I need to have some explicit negotiation > between libxc and libxl about sharing ownership of the input fd. This > is going to require a new record in the format, and I currently drafting > a patch or two which should help in this regard. > > My view for the eventual stream looks something like this (time going > downwards): > > libxc writes: libxl writes: > > Image Header > Domain Header > start_of_stream() > start_of_checkpoint() ctx->save.callbacks->suspend() this callback suspend the primary guest and then calls Remus devices postsuspend callbacks to buffer the network pkts etc. > end_of_checkpoint() > Checkpoint record ctx->save.callbacks->postcopy() this callback should not be omitted, it do some necessary work before resume primary (such as call Remus devices preresume callbacks to ensure the disk data is consistent) and then resume the primary guest. I think this callback should be renamed to ctx->save.callbacks->resume(). > ctx->save.callbacks->checkpoint() > libxl qemu record Maybe we should add another callback to send qemu record instead of using checkpoint callback. We can call it ctx->save.callbacks->save_qemu() Then in checkpoint callback, we only call remus devices commit callbacks( which will release the network buffer etc...) then decide whether we need to do another checkpoint or quit checkpointed stream. With Remus, checkpoint callback only wait for 200ms(can be specified by -i) then return. With COLO, checkpoint callback will ask COLO proxy if we need to do a checkpoint, will return when COLO proxy module indicate a checkpoint is needed. > ... > libxl end-of-checkpoint record > ctx->save.callbacks->checkpoint() returns > start_of_checkpoint() ctx->save.callbacks->suspend() > > end_of_checkpoint() > Checkpoint record > etc... > > This will eventually allow both libxc and libxl to send checkpoint data > (and by the looks of it, remove the need for postcopy()). With this > libxc/remus work it is fine to use XG_LIBXL_HVM_COMPAT to cover the > current qemu situation, but I would prefer not to be also retrofitting > libxc checkpoint records when doing the libxl/migv2 work. > > Does this look plausible in for Remus (and eventually COLO) support? With comments above, I would suggest the save flow as below: libxc writes: libxl writes: live migration: Image Header Domain Header start_of_stream() start_of_checkpoint() ctx->save.callbacks->suspend() end_of_checkpoint() if ( checkpointd ) End of Checkpoint record /*If resotre side receives this record, input fd should be handed to libxl*/ else goto end loop of checkpointed stream: ctx->save.callbacks->resume() ctx->save.callbacks->save_qemu() libxl qemu record ... libxl end-of-checkpoint record /*If resotre side receives this record, input fd should be handed to libxc*/ ctx->save.callbacks->save_qemu() returns ctx->save.callbacks->checkpoint() start_of_checkpoint() ctx->save.callbacks->suspend() end_of_checkpoint() End of Checkpoint record goto 'loop of checkpointed stream' end: END record /*If resotre side receives this record, input fd should be handed to libxl*/ In order to keep it simple, we can keep the current ctx->save.callbacks->checkpoint() as it is, which do the save_qemu thing, call Remus devices commit callbacks and then decide whether we need a checkpoint. We can also combine the ctx->save.callbacks->resume() with ctx->save.callbacks->checkpoint(), with only one checkpoint() callback, we do the following things: - Call Remus devices preresume callbacks - Resume the primary - Save qemu records - Call Remus devices commit callbacks - Decide whether we need a checkpoint Overall, there are 3 options for the save flow: 1. keep the current callbacks, rename postcopy() to resume() 2. split the checkpoint() callback to save_qemu() and checkpoint() 3. combine the current postcopy() and checkpoint() Which one do you think is the best? > > ~Andrew > . > -- Thanks, Yang.