[sheepdog-users] sheepdog recovery question

[Hitoshi Mitake]

At Sat, 31 May 2014 07:49:25 +0800,
=?gb18030?B?t8k=?= wrote:
> 
> [1  <multipart/alternative (8bit)>]
> [1.1  <text/plain; gb18030 (base64)>]
> hello, I have running a cluster with 3 nodes. and I store more than 3TB data in my cluster,
> there are 50 VMs running on the cluster.
> but one node have crash yestoday, when I restart the crashed sheep, my cluster going to recovery, then, many VMs are very slow, I check the disk idle time of the vm, it often shows 0, I think many IO are block when recovery.  then I check the disk, iostat and the bandwidth of all sheep nodes, there are all normal .   I think the nodes hardware are very well,the reason of vms slow is caused by sheepdog , how to solve this issue? ------------------ 原始邮件 ------------------
> 发件人: "我自己的邮箱"<duron800 at qq.com>
> 发送时间: 2014年5月13日(星期二) 下午3:18
> 收件人: "sheepdog-users"<sheepdog-users at lists.wpkg.org>;
> 主题: sheepdog about clone
> 
> 
> hello, I have a question for clone a sheepdong image.
> my clone steps are:
> 1. I create a image named myimg
> 2. I write some files to  myimg
> 3. I create a snapshot named 'newest', dog vdi snapshot -s newest myimg
> 4. I ran 'dog vdi clone -s newest myimg myimg2'
> 5. I ran 'dog vdi delete myimg'
> 6. I attach the myimg2 to QEMU, but I can't found the files I wrote at step 2.
> 
> what happen ? if I delete the source image 'myimg', it will be make data lost ?

This problem can happen when the guest OS doesn't flash its page
cache. You can avoid this problem by using qemu-ga. The detailed
instruction is here:
https://github.com/sheepdog/sheepdog/wiki/How-to-make-a-live-snapshot-properly

Thanks,
Hitoshi

> [1.2  <text/html; gb18030 (base64)>]
> 
> [2  <text/plain; us-ascii (7bit)>]
> -- 
> sheepdog-users mailing lists
> sheepdog-users at lists.wpkg.org
> http://lists.wpkg.org/mailman/listinfo/sheepdog-users