Thank you for your replay.<div>I use the latest version 0.2.4. <br>I create 8 processes on different port in one server.</div><div>If I don't use the preallocation mode,it works well,but i have not done a lot of tests.</div>
<div><br></div><div>Thanks,</div><div>Dong Wu<br><br><div class="gmail_quote">在 2012年1月6日 下午8:45，MORITA Kazutaka <span dir="ltr"><<a href="mailto:morita.kazutaka@lab.ntt.co.jp">morita.kazutaka@lab.ntt.co.jp</a>></span>写道：<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">At Fri, 6 Jan 2012 20:07:15 +0800,<br>
<div class="im">dong wu wrote:<br>
> Hello,<br>
> when i create a preallocation vdi,during the creating,one node left and<br>
> then create error!<br>
> I do like this:(I use a server with 8disks)<br>
> [root@sheepdog ~]# collie node list<br>
> Idx - Host:Port Vnodes Zone<br>
> ---------------------------------------------<br>
> * 0 - <a href="http://192.168.130.60:7000" target="_blank">192.168.130.60:7000</a> 64 1015195840<br>
> 1 - <a href="http://192.168.130.60:7001" target="_blank">192.168.130.60:7001</a> 64 1015195840<br>
> 2 - <a href="http://192.168.130.60:7002" target="_blank">192.168.130.60:7002</a> 64 1015195840<br>
> 3 - <a href="http://192.168.130.60:7003" target="_blank">192.168.130.60:7003</a> 64 1015195840<br>
> 4 - <a href="http://192.168.130.60:7004" target="_blank">192.168.130.60:7004</a> 64 1015195840<br>
> 5 - <a href="http://192.168.130.60:7005" target="_blank">192.168.130.60:7005</a> 64 1015195840<br>
> 6 - <a href="http://192.168.130.60:7006" target="_blank">192.168.130.60:7006</a> 64 1015195840<br>
> 7 - <a href="http://192.168.130.60:7007" target="_blank">192.168.130.60:7007</a> 64 1015195840<br>
<br>
</div>It seems that your Sheepdog version is not latest. Can you update it<br>
on all nodes?<br>
<div class="im"><br>
> [root@sheepdog ~]# qemu-img create sheepdog:Alice -o preallocation=data 100G<br>
> Formatting 'sheepdog:Alice', fmt=raw size=107374182400 preallocation='data'<br>
> qemu-img: No object found<br>
><br>
> qemu-img: sheepdog:Alice: error while creating raw: Operation not permitted<br>
> when create error,check the node list as follows:<br>
<br>
</div>If you don't use the preallocation mode, does Sheepdog work well?<br>
<div><div class="h5"><br>
> [root@sheepdog ~]# collie node list<br>
> Idx - Host:Port Vnodes Zone<br>
> ---------------------------------------------<br>
> * 1 - <a href="http://192.168.130.60:7001" target="_blank">192.168.130.60:7001</a> 64 1015195840<br>
> 2 - <a href="http://192.168.130.60:7002" target="_blank">192.168.130.60:7002</a> 64 1015195840<br>
> 3 - <a href="http://192.168.130.60:7003" target="_blank">192.168.130.60:7003</a> 64 1015195840<br>
> 4 - <a href="http://192.168.130.60:7004" target="_blank">192.168.130.60:7004</a> 64 1015195840<br>
> 5 - <a href="http://192.168.130.60:7005" target="_blank">192.168.130.60:7005</a> 64 1015195840<br>
> 6 - <a href="http://192.168.130.60:7006" target="_blank">192.168.130.60:7006</a> 64 1015195840<br>
> 7 - <a href="http://192.168.130.60:7007" target="_blank">192.168.130.60:7007</a> 64 1015195840<br>
> I have tried this several times,always the master node left.<br>
> and sheep.log as follows when fails:<br>
> ........<br>
> Jan 06 02:44:31 store_queue_request(817) 3, 3, 28489a00000083 , 1<br>
> Jan 06 02:44:31 store_queue_request_local(571) 3, 28489a00000083 , 1<br>
> Jan 06 02:46:12 sd_confchg(1872) confchg nodeid 3282a8c0<br>
> Jan 06 02:46:12 sd_confchg(1874) 6 1 0<br>
> Jan 06 02:46:12 sd_confchg(1877) [0] node_id: 3282a8c0, pid: 3076<br>
> Jan 06 02:46:12 sd_confchg(1877) [1] node_id: 3282a8c0, pid: 3092<br>
> Jan 06 02:46:12 sd_confchg(1877) [2] node_id: 3282a8c0, pid: 3108<br>
> Jan 06 02:46:12 sd_confchg(1877) [3] node_id: 3282a8c0, pid: 3124<br>
> Jan 06 02:46:12 sd_confchg(1877) [4] node_id: 3282a8c0, pid: 3140<br>
> Jan 06 02:46:12 sd_confchg(1877) [5] node_id: 3282a8c0, pid: 3156<br>
> Jan 06 02:46:12 sd_confchg(1891) allow new confchg, 0x26e3f20<br>
> Jan 06 02:46:12 start_cpg_event_work(1716) 0 0<br>
> ........<br>
> at times,I create succefully,but when I use iscsi to mount on one linux<br>
> host,then use iometer to test the performance<br>
> during the test,this error occurs again,the master node left,even worse,all<br>
> the other nodes left one by one.<br>
><br>
> If I create a preallocation vdi not that big(eg. 15G),i can create<br>
> successfully,<br>
> but the performance i test using iometer is very slow<br>
> 4k %100read 100%sequence 2.897764M/s<br>
> 4k %100write 100%sequence 0.089305M/s<br>
> 1M %100read 100%sequence 29.83363M/s<br>
> 1M %100write 100%sequence 10.05084M/s<br>
> 4M %100read 100%sequence 50.71799M/s<br>
> 4M %100write 100%sequence 9.699951M/s<br>
<br>
</div></div>Currently, Sheepdog writes data with O_SYNC always, so write<br>
performance is bad. It would improve if we implement write-cache<br>
support.<br>
<br>
Thanks,<br>
<br>
Kazutaka<br>
</blockquote></div><br></div>