Thanks for the update Gregoire, Would you be able to check the vhd-util versions in your working vs not-working scenarios?
Regards, Dag Sonstebo Cloud Architect ShapeBlue On 06/03/2018, 22:11, "Grégoire Lamodière" <g.lamodi...@dimsi.fr> wrote: Hi Dag, All, I spent some time working on this matter, and here are the results of my tests. It might be usefull in case anyone has to restore instances from nfs store. I think I were facing 2 issues. 1/ You are right, in case vm crashed (ie was running at the time of network crash), you may assume around 50% luck to restore it easily. 2/ vhd-util version used in the xenserver seems to be the key of the process. I made the following tries : - take vhd from cs 4.9.3 / xenserver 7.0 - download VHD-UTIL from install guide url - Import it to cs 4.11 / xenserver 7.0 => not working - Upgrade the xenserver to 7.2 => import still not working - Import directly to xenserver 7.0 / 7.2 => working 100% - make exactly the same, copying vhd-util from original cs 4.9.3 management server to the new one - Import it to cs 4.11 / xenserver 7.0 => working - Upgrade the xenserver to 7.2 => import working So from what I can see : Vdi-import always works in a Xenserver perspective, but in some cases vm won't start properly if they were inconsistent. When vhd is downloaded as a template in cs, the errors reported were mostly post-download errors, ex. In the ssvm script : vhd-util set -n ${tmpltimg2} -f "hidden" -v "0" > /dev/null rollback_if_needed $tmpltfs $? "vhd remove $tmpltimg2 hidden failed\n That’s all for my feedback. In case anyone needs info or scripts, I'll be happy to share. Now it's time to continue my work moving to KVM, but that is another story. Regards Grégoire --- Grégoire Lamodière T/ + 33 6 76 27 03 31 F/ + 33 1 75 43 89 71 -----Message d'origine----- De : Grégoire Lamodière [mailto:g.lamodi...@dimsi.fr] Envoyé : vendredi 23 février 2018 21:10 À : users@cloudstack.apache.org Objet : RE: VHD import Hi Dag, I am still trying to understand, but some instances crashed where some others shutdown properly. So far, I can confirm that all vhd (after coalesce) can be imported to Xen, and most of them fail to import as VHD. Some failure imports are successful after the Xen import step, but some still fail after the xe vhd-export (xen vm is working fine) The strange thing is all vhd sound good (all steps of the cs script pass, but the process results in cs errors). This is not a big deal, as we are moving to KVM, but I hope we won't have the same issues with kvm nfs based volumes. I'll work on this a bit this weekend, so I can make a proper feedback. Anyway, once again, thanks for your blog post, clear and usefull, as usual. Grégoire --- Grégoire Lamodière T/ + 33 6 76 27 03 31 F/ + 33 1 75 43 89 71 -----Message d'origine----- De : Dag Sonstebo [mailto:dag.sonst...@shapeblue.com] Envoyé : vendredi 23 février 2018 20:48 À : users@cloudstack.apache.org Objet : Re: VHD import Hi Gregoire, One thing just struck me – what was the status of the VMs you were exporting from NFS? Were they cleanly shut down or did they simply crash when the hardware failed? If so could you have an issue where the disk you are initially exporting isn’t consistent, hence import fails? Regards, Dag Sonstebo Cloud Architect ShapeBlue On 22/02/2018, 09:56, "Dag Sonstebo" <dag.sonst...@shapeblue.com> wrote: Hi Gregoire, Glad the blog post is of use. It’s a while since I wrote it so I had to re-read it myself. I have not come across this problem – but as you can probably guess we luckily don’t have to do this recovery procedure very often. I can only assume same as yourself that the vhd-util coalesce using the downloaded vhd-util binary must give a slightly different format header which the import in CloudStack doesn’t like but XS is quite happy about. Also keep in mind that the vhd-util from http://download.cloud.com.s3.amazonaws.com/tools/vhd-util is slightly different than the one you find on the XenServers, so this may also be a factor. Please let us know how you get on with your investigation – if you find the root cause let me know and I’ll add it as a subnote to the original article (don’t worry you’ll get the credit ( ). Regards, Dag Sonstebo Cloud Architect ShapeBlue On 22/02/2018, 00:10, "Grégoire Lamodière" <g.lamodi...@dimsi.fr> wrote: Dear All, I recently lost a Xen 7 pool managed by CS 4.11 (hardware issue). I tried to apply Dag's excellent article (http://www.shapeblue.com/recovery-of-vms-to-new-cloudstack-instance/), and found a strange behavior : - If I directly export the VHD from old NFS primary storage (after coalesce vhd with vhd-util) and import to CS using import Template, they always fail, with 2 types of error (mainly Failed post download script: vhd remove /mnt/SecStorage/2da072e7-a6fe-3e39-b07e-77ec4f34fd49/template/tmpl/2/258/dnld5954086457185109807tmp_ hidden failed). - If I use the same VHD, import-it to a xen server (xe vdi-import), and then export it (xe vdi-export), I can successfully import it to CS I made tries with Windows instances, from 15 to 100 Gb, and always the same result. And I also made the same to a CS 4.7 with same results. A quick look inside createtmplt.sh did not show anything relevant. I guess the header / footer might have some differences, I'll check this tomorrow. Is anyone aware of this ? Is there any solution to avoid the intermediate vdi-import / export. I will dig a bit more to find a proper solution, as it sounds interesting to get a disaster migration script working fine. Regards. --- Grégoire Lamodière T/ + 33 6 76 27 03 31 F/ + 33 1 75 43 89 71 dag.sonst...@shapeblue.com www.shapeblue.com 53 Chandos Place, Covent Garden, London WC2N 4HSUK @shapeblue dag.sonst...@shapeblue.com www.shapeblue.com 53 Chandos Place, Covent Garden, London WC2N 4HSUK @shapeblue dag.sonst...@shapeblue.com www.shapeblue.com 53 Chandos Place, Covent Garden, London WC2N 4HSUK @shapeblue