My understand of how snapshot works (XenServer as hypervisor) in CloudStack is:
1. User takes a snapshot of volume either manually or by schedule 2. CloudStack does API call to XS for taking a snapshot of the particular VHD - this happens on primary storage 3. CloudStack use SSVM to copy the newly created snapshot (VHD) from primary to secondary storage 4. CloudStack make API call to coalesce snapshot chain for the volume on primary storage This way CloudStack can keep snapshot chain length minimum on primary storage. It should never grow near the XenServer limit of 30 snapshots. As of VHD snapshot chain on secondary storage, it is different from primary storage's. There is a global setting 'snapshot.delta.max' (default to 16) to determine how often should CS start a new chain with full copy of VHD. Above is what my understanding is. If it's not accurate please let me know. And my problem is some VHD's snapshot chain does hit to the limit of 30 on XenServer. Regards, Yong -----Original Message----- From: Kirk Jantzer [mailto:[email protected]] Sent: Monday, 15 July 2013 10:47 AM To: Cloudstack users mailing list Subject: Re: The snapshot chain is too long Don't think of CS snapshots the same way you do as VMWare snapshots. CloudStack snapshots are WHOLE COPY BACKUPS of the instance. Also, in regards to snapshot chains, I know VMWare has a default setting per VM of 32 snapshots. If you exceed this, it will stop you from creating a new one. You can, however, modify that setting so that you can create more. I am not sure if XenServer has the same type of setting. On Sun, Jul 14, 2013 at 8:43 PM, Yong Chen <[email protected]> wrote: > Hi, > > I'm using CS 4.0.1 and XS 6.0.2. There is a problem with taking > snapshots of instances. > > Some of VMs have scheduled snapshots (e.g. weekly, monthly) so > eventually there will be quite some number of snapshots on secondary > storage. This is fine as I read CloudStack should copy snapshots to > secondary storage and then coalesce snapshots on primary storage so > can keep the chain not too long (like 3-4 child normally). > > However I found after snapshot been taken for a while it starts > failing due to error logged on CS Management log > 'SR_BACKEND_FAILURE_109The snapshot chain is too long'. I verified it > by checking with vhd-util and it does show the affected VHD has > exactly 30 snapshots in chain which is the limit for XenServer. > > I tried to search for info why the chain can grow to 30 and only found > this article for CloudStack 2.2 > http://support.citrix.com/article/CTX133470 seems to be related. It > mentioned there maybe problem preventing XenServer from cleaning up > snapshots. However I can't find any more error info other than CS > Management log 'SR_BACKEND_FAILURE_109The snapshot chain is too long'. > > What can cause XenServer or CloudStack not cleaning up snapshots from > primary storage? Thanks in advance. > > Yong > -- Regards, Kirk Jantzer c: (678) 561-5475 http://about.met/kirkjantzer
