On 07/25/2013 09:51 PM, Liu Yuan wrote: > On Thu, Jul 25, 2013 at 07:35:10PM +0000, Andrew J. Hobbs wrote: >> Just finished running a successful snapshot to the SMB share. Thank >> you! Total disk usage of the snapshot is 206G, total virtual image size >> is 270G, with collie node info listing 98G used. I'm going to be >> deploying more virtual machines on these nodes over the next few days, >> and adding at least 2 more nodes to the cluster. >> > That is strange. Suppose you have 2 copies, and then we have to backup 98/2=49G, > then the cluster snapshot will take less than 49G. But as you reported, it > use much more than that! I have no idea why it happend. > >> Not sure how to exactly determine amount of deduplication at this >> point. The issue is a bit muddled as my data volumes are btrfs with >> compress=lzo set, and my images are stored in sheepdog in qcow2 format. >> So far the performance has been quite satisfactory in spite of this. > Any reason you use qcow2? raw(sheedpog default) is best performance and enjoy > full features such as snapshot, clone, thin provision, etc. Space is at a premium at this point, and I've been virtualizing existing boxes. I've been restoring backups of physical boxes into virtual boxes on my desktop box, then loading the resulting qcow2 image into the cluster. No particular reason for qcow2 besides habit and thin provisioning. I would like to convert one to a raw format and compare performance. > Thanks > Yuan > Going to be reformatting the cluster today and restoring from the cluster snapshot. Existing cluster is with 2 copies, going to go to three. Next week I'll be adding in more nodes. -------------- next part -------------- A non-text attachment was scrubbed... Name: ajhobbs.vcf Type: text/x-vcard Size: 353 bytes Desc: ajhobbs.vcf URL: <http://lists.wpkg.org/pipermail/sheepdog-users/attachments/20130726/b4b2c0a9/attachment.vcf> |