Hi Tim, Thank you very much for your reply and your offer to mentor our efforts (you are welcome for that). Yes we have enabled local storage via management-server but still we didn’t manage to get the local storage of the slave node. Our solution to that was exactly what you mentioned: We divided our cluster (1 cluster -> 4 compute nodes) into tiny sub-clusters where each of them has one compute node (4 clusters -> 4 compute nodes). So every compute node is a pool-master, thus we can retrieve its local storage.
It’s not an ideal solution and it’s not how things should be but still works and we don’t see any drawback on that (but we don’t see any benefit either, except getting the local storage). So soon we will deliver our platform based on Cloudstack 4.6.1 and XenServer Dundee and we will keep going with improvements and automation. I am more than happy to keep the community up-to-date with our progress and findings. But I would like to ask if that is a bug on Cloudstack in order to report it officially, or people started working on improving the XenServer module and they considered the issue already. P.S: Another lesson learned yesterday: MySQL 5.7.9 is incompatible currently with CS 4.6.1 Cheers Stavros ---------------------------- Stavros Konstantaras Science faculty Research IT support (FEIOG) University of Amsterdam, Science Park 904, 1098 XH Fingerprint: E5E5 9B19 D1CD 88CD 4763 3465 A8DC 7C92 330F D59A > On 04 Dec 2015, at 21:40, Tim Mackey <tmac...@gmail.com> wrote: > > Stavros, > > Have you enabled local storage within CloudStack ( > http://xmodulo.com/how-to-use-local-storage-for-cloudstack-vms.html)? If > so, its possible local storage implies a CloudStack cluster with only one > host (i.e. you need to break the XenServer pool into independent hosts with > each being in a CloudStack cluster). I'd need to dig fairly deep in the > code to confirm that theory. > > To your question about XenServer Dundee (what you're seeing as 6.6.90), > since it hasn't yet been released by Citrix the visual artifacts (wrong > version number) are to be expected. Having just checked the database code, > it doesn't look like the work has yet been done to even provide > experimental support for Dundee. If you would like to remedy that, I think > the community would welcome your contributions, and I could help mentor in > the effort. Just let me know. > > -tim > > On Fri, Dec 4, 2015 at 5:37 AM, Stavros Konstantaras <s.konstanta...@uva.nl> > wrote: > >> Hi everyone, >> >> At my organisation we tried to build a tiny cluster (1 head, 2 compute >> nodes) with CloudStack 4.6 and XenServer 6.6.90 for experimental reasons. >> As we use local storage for deploying VMs, we discovered that Cloudstack is >> not able to recognise the local storage pool of the slave compute node. >> >> We tried reversing the master node (make the slave compute node -> >> poolmaster and the poolmaster -> slave) but the result was the same. We >> tried Cloudstack 4.5.2 and again problem didn’t solve. We checked the >> database of the poolmaster and it was totally updated and able to detect >> the local storage of the slave node. On Cloudstack-management interface, >> the XenServer 6.6 hosts appear as XenServer 6.5 hosts. In the meanwhile, we >> are totally able to deploy VMs in poolmaster and make use of the local >> storage. >> >> Did any of you face this issue and if yes, did you manage to solve it? I >> suspect that there are some changes in the API of Xenserver 6.6 which CS >> does not support currently. >> >> Best Regards >> Stavros >> >> ---------------------------- >> Stavros Konstantaras >> Science faculty Research IT support (FEIOG) >> University of Amsterdam, Science Park 904, 1098 XH >> >> Fingerprint: E5E5 9B19 D1CD 88CD 4763 3465 A8DC 7C92 330F D59A >> >>
