>>> When disk is file based , it is possible. >>> When its normal disk and have some file system like ZFS or volume >>> manager which supports snapshot , it's possible.Otherwise its not >>> When its LUN from filer , if filer has snapshot capabilities , it is >>> supported. otherwise not. > >> You are correct. For disk image snapshot, this depends on the capability >> of the backend storage where the disk image is stored. The easiest solution >> is to store disk images with ZFS and you will have quick snapshot features. >> > > hmm .. wondering when the disk is actually a file. Even in that case > LDom does not support snapshot. I still need to use some other tool > to get the snapshot of the 'virtual disk' file ? I understand now > this will be the case for normal disk or LUN device exported from > filer. But for file ?
To snapshot a file, you can just copy it, or use any snapshot feature provided by the filesystem. If the file is on a ZFS filesystem then you can snapshot the ZFS filesystem. >> Note that when you snapshot a disk image, you only take a snapshot of >> the virtual disk, this does not include the domain configuration or the >> memory state. >> > > Ok .. got it. So we do not really snaps the Ldom instance per say but > only the devices that it might be using. Yes, you can only snapshot the virtual disks. >>> Also my question would be, does LDom admin tool takes care of these >>> snapshot work in the back end? Or it is the user's responsibility >>> to manage all these different snapshots. >> This has to be done manually. >> > > Are there are any open source tools known which can integrate > with(work with) LDom to create snapshot if devices can be snapped > using some some technology.(file NetApp filer can be snapped, some > volume managers can take snapshot). I know this has to happen > independently but just wondering if anybody has already solved this > problem. I am not aware of any such tool, the problem is there's a lot of different technologies and you would need to be compatible with every different products and vendors. That's not a simple thing. >>> 3: What is the preferred way to take LDom snapshots in community ? >> Use ZFS to store disk images (either on ZFS filesystem or volume), then >> use 'zfs snapshot/clone'. This is frequently used to clone an existing >> domain to create a new one. >> > > Am I correct in assuming 'Even in case of ZFS , snapshot of disk will > be to be taken manually and will not be controlled by LDom admin util > ' ? Yes, you have to directly use the 'zfs snapshot' command, there's no integration with the domain manager. >>> 4:Does Ldom have suspend and resume support at machine level ? >> There's a suspend/resume feature but it can not be used directly. It is >> used by warm migration. During warm migration, a domain is going to be >> suspended, its memory and cpu state is going to be transfered to another >> platform where it will be resumed. >> But there's currently no option to use this feature to suspend a >> domain to disk. >> > > You mean in warm migration cpu state and memory state is transferred > directly to target LDom instance and there is no way for now to dump > that info to disk. hmm Is there any plan to add such support in near > future ? No plan for that at the moment, but let us know if you have requirements for such a feature. Actually you can dump states to files, but that's only for internal use to debug migration problems but this requires a debug version of the domain manager. >>> 5: Is there any way for third party application to access these >>> snapshots and memory state ? Is there any too/library which provides >>> interface for accessing such snapshots and other machine state (e.g >>> memory in case of hibernation). Any links ? > > So if its ZFS or some volume manager , corresponding tools are the > only thing we need to use to create snapshot and back them up > somewhere else and there is no generic api/tools which cover all these > different technologies(filer/ZFS/file based) under on hood rht ? The problem is that way to do snapshot is very different from one disk backend to another. Basically you would have to identify the virtual disk backend: ZFS, SVM, VxVM, physical disk, NFS, storage array (which type?) and then invoke the appropriate snapshot method. Especially if you have a storage array, this can be very different when using EMC, Hitachi or Sun storage (even for different arrays from the same vendor). alex.
