[sheepdog] [PATCH 1/2] collie: add command 'collie vdi rollback'

[levin li]

On 08/31/2012 04:25 PM, MORITA Kazutaka wrote:
> Here is a draft patch which implements my example but doesn't support
> resetting clone vdis.  I hope it would help you.
> 
> ==
> commit 138511431d4f87642572b0e2a51d869240b5ac83
> Author: MORITA Kazutaka <morita.kazutaka at lab.ntt.co.jp>
> Date:   Fri Aug 31 17:21:38 2012 +0900
> 
>     [tmp] collie: add vdi rollback support
>     
>     Signed-off-by: MORITA Kazutaka <morita.kazutaka at lab.ntt.co.jp>
> 
> diff --git a/collie/vdi.c b/collie/vdi.c
> index 966fb9f..39e04a7 100644
> --- a/collie/vdi.c
> +++ b/collie/vdi.c
> @@ -752,6 +752,37 @@ static int vdi_delete(int argc, char **argv)
>  	return EXIT_SUCCESS;
>  }
>  
> +static int vdi_rollback(int argc, char **argv)
> +{
> +	char *vdiname = argv[optind++];
> +	uint32_t base_vid;
> +	int ret;
> +	char buf[SD_INODE_HEADER_SIZE];
> +	struct sheepdog_inode *inode = (struct sheepdog_inode *)buf;
> +
> +	if (!vdi_cmd_data.snapshot_id && !vdi_cmd_data.snapshot_tag[0]) {
> +		fprintf(stderr, "Please specify the '-s' option\n");
> +		return EXIT_USAGE;
> +	}
> +
> +	ret = find_vdi_name(vdiname, vdi_cmd_data.snapshot_id,
> +			    vdi_cmd_data.snapshot_tag, &base_vid, 0);
> +	if (ret < 0) {
> +		fprintf(stderr, "Failed to open VDI %s\n", vdiname);
> +		return EXIT_FAILURE;
> +	}
> +
> +	inode = xmalloc(sizeof(*inode));
> +	ret = sd_read_object(vid_to_vdi_oid(base_vid), inode, SD_INODE_SIZE, 0);
> +	if (ret != SD_RES_SUCCESS) {
> +		fprintf(stderr, "Failed to read an inode\n");
> +		return EXIT_FAILURE;
> +	}
> +
> +	return do_vdi_create(vdiname, inode->vdi_size, base_vid, NULL,
> +			     inode->snap_id, vdi_cmd_data.nr_copies);
> +}
> +
>  static int vdi_object(int argc, char **argv)
>  {
>  	char *vdiname = argv[optind];
> @@ -1537,6 +1568,9 @@ static struct subcommand vdi_cmd[] = {
>  	{"delete", "<vdiname>", "saph", "delete an image",
>  	 NULL, SUBCMD_FLAG_NEED_THIRD_ARG,
>  	 vdi_delete, vdi_options},
> +	{"rollback", "<vdiname>", "saph", "rollback to a snapshot",
> +	 NULL, SUBCMD_FLAG_NEED_THIRD_ARG,
> +	 vdi_rollback, vdi_options},
>  	{"list", "[vdiname]", "aprh", "list images",
>  	 NULL, 0, vdi_list, vdi_options},
>  	{"tree", NULL, "aph", "show images in tree view format",
> 

With your patch I got the following result:

$ collie/collie vdi list
  Name        Id    Size    Used  Shared    Creation time   VDI id  Copies  Tag
s base         1   20 MB   20 MB  0.0 MB 2012-08-31 16:58   54c278     3           tag
s base         2   20 MB  0.0 MB   20 MB 2012-08-31 16:58   54c279     3              
  base         3   20 MB  0.0 MB   20 MB 2012-08-31 16:59   54c27a     3              
c clone        1   20 MB  0.0 MB   20 MB 2012-08-31 16:58   72a1e2     3              

$ collie/collie vdi tree
base---[2012-08-31 16:58]---[2012-08-31 16:58]---(you are here)
clone---(you are here)
$ collie/collie vdi rollback -s 2 base
$ collie/collie vdi tree
base---[2012-08-31 16:58]---[2012-08-31 16:58]-+-[2012-08-31 16:59]
                                               `-(you are here)
clone---(you are here)
$ collie/collie vdi rollback -s 1 base
$ collie/collie vdi tree
base---[2012-08-31 16:58]-+-[2012-08-31 16:58]-+-[2012-08-31 16:59]
                          |                    `-[2012-08-31 17:00]
                          `-(you are here)

I still don't understand what's the use of your rollback of a snapshot, it seems
it just move the current VDI to the specified snapshot VDI and created a new one,
but what's the use of the current VDI ?

In my scenario, we clone a VDI from a snapshot, and the cloned VDI is modified after
VM running, but we want to reset the VDI back to the initial state, I think this is
useful for me and many users, but I don't quite understand the use of your rollback
scenario, maybe I can go back to rename the command as *'reset'* to just reset a cloned
VDI, what do you think about it?

thanks,

levin