[sheepdog] 答复: Could cluster-wide snapshot be restored into different cluster with different number nodes?

[Ruoyu]

On 2014年09月22日 15:31, ivanzhu wrote:
>
> Hi, Ruoyu, thanks for  your reply.
>
> Yes, I have found there are some fix on stable 0.8, but we still use 
> stable 0.8.2. Update will be risky, so we need some time.
>
> I still have some inline questions, can you help me again?
>
> *发件人:*Ruoyu [mailto:liangry at ucweb.com]
> *发送时间:*2014年9月22日14:01
> *收件人:*朱波; sheepdog
> *主题:*Re: [sheepdog] Could cluster-wide snapshot be restored into 
> different cluster with different number nodes?
>
> On 2014年09月22日13:36, 朱波wrote:
>
>     Hi, folks
>
>        I'm a new sheepdog user, I have run sheepdog on 15 sheepdog
>     nodes. And do cluster-wide snapshot for this cluster.
>
>     Recently, I restored the snapshot into a new cluster with 7 nodes.
>     But it failed. the command "dog vdi list" would show me some
>     object can't be found. And currently we don't have more nodes and
>     time to verify whether it would be successful when restore it into
>     a new cluster with 15 sheepdog nodes.
>
>     So I have two questions want to check with you guru:
>
>     1. Could cluster-wide snapshot be restored  into different
>      cluster with different number node?
>
> Yes. Cluster-wide snapshot is de-duplicated. It can be restored into 
> different cluster with different number of nodes and with different 
> redundancy level.
>
> Actually, I find there is aconstraint for cluster-wide snapshot in the link:  https://www.mail-archive.com/sheepdog@lists.wpkg.org/msg06209.html  (at the bottom of page).
> I’m not sure the cluster topology changes means the cluster node number, if you can help explain this, it will be greate.
>   
It is written in 2012. Maybe it is not the same as today.
Yuan, what's your opinion? It seems it's you answered the mail and 
mentioned cluster topology.
>
>     ‍
>
>     2. If one cluster-wide snapshot is failed(after restored,  "dog
>     vdi list" will show some object can't be found), would another new
>     cluster-wide snapshot based on the failed one be successful? in
>     other words, the first tag of cluster-wide snapshot is failed, can
>     the second tag be successful? Are they independent?
>
> What is your sheep's version? And, how you produce the failure (some 
> objects not found)? Maybe it is relevant to a bug which is fixed recently.
> https://github.com/sheepdog/sheepdog/commit/7e236cee8951aa749b84e9b66fbb86385a92b450
>
> the operation step:
>
> 1.dog cluster snapshot save snapshot_1  BACKUP_DIR -a 192.168.1.102 
>  -p 7000  (snapshot the cluster A)
>
> 2.dog cluster snapshot load snapshot_1 BACKUP_DIR –a 177.xx.xx.xx   -p 
> 7000  (restore into the cluster B)
>
> 3.on the cluster B. dog vdi list
>
> it will display xxxx object can’t be found. (I’m not sure why this 
> happen).
>
> 4.dog cluster snapshot save snapshot_2  BACKUP_DIR -a 192.168.1.102 
>  -p 7000  (snapshot the cluster A)
>
> *Can the snapshot_2  be successful? Or It must be failed like the 
> snapshot_1?*
>
I tried your operation steps with sheepdog v0.8.2, but no error message 
found in my environment while executing dog vdi list.
Could yo please check your sheep's log? Maybe it is helpful for analysis.
>
>     Anybody can help me? Any response will be appreciated.‍
>
>     Thanks & Regards
>
>     Ivan
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.wpkg.org/pipermail/sheepdog/attachments/20140922/f49b9bb7/attachment-0004.html>