These forums are locked and archived, but all topics have been migrated to the new forum. You can search for this topic on the new forum: Search for Snapshots of VMs - Interesting for you? Please reply! on the new forum.
Web Hosting and Cloud Computing Control Panels
I would be interested, yes!
Count me in. Snapshots where really good to have when I was using VMware ESXi. Applying updates to production systems was less scary!
I think it's a great idea. And I hope it would be usable for backups - backup of snapshot should be consistent because memory content is part of the snapshot.
I think you can already easily take snapshots if you use LVM.
Yes, that's true but I think it's just "snapshot" of HDD, there is no memory content. So it's just like you press reset button on computer when you restore the snapshot. What I am concerned about the most is the database running in VM - there could be completed transactions in cache, not stored on disk and after restore you have inconsistent database.
Even though I don't use Cloudmin (VMWare ESXi here), I can say from my perspective that snapshots is a quite vital feature for me. I use them regularly on ESXi before making critical changes to production systems. Nothing is easier than rolling back from problems after an update when you have a snapshot. :)
In ESXi, I can choose per snapshot if I wish to include the whole system state (HDD, memory contents, virtual device status etc.) or just the HDD contents. I can also - through the VMWare Tools - have it quiesce the guest file system (i.e. flush buffers and redirect further writes to a temporary area) before creating the snapshot, to improve data consistency.
Snapshots in ESXi are "copy-on-write". Creating a snapshot is very fast, more or less instantaneous. After a snapshot is created, the main HDD files are not changed anymore, and writes to the file system are executed in a new HDD file which only contains the changes. That means that while snapshots exist, disk IO performance is slightly degraded, since for the first write access to each block that block must first be read from the main file, then modified and written to the snapshot file.
Deleting a snapshot means copying back the changes in the snapshot file to the main HDD file, and can take some time depending on snapshot size (here meaning the amount of data that was modified since the snapshot was taken).
So: If I were to decide to switch away from VMWare to KVM/Xen, having a snapshot feature would be a "must-have" for me.
I think it's a vitual feature which CloudMin lacks dearly.
It would be great if each user could create a snapshot (and have the ability to restore from it) before doing any major updates. We had a case recently where one of our DNS servers which had a much older version of BIND, was upgraded to the latest one and we were forced to format & reinstall from the scratch the DNS server due a major problem this upgrade caused. After spending about 9 days trying to sort out the problem, we decided to format & reinstall - which was quicker than spending more time trying to figure out what went wrong. This also meant that I lost a lot of zone files that was added / updated in those 9 days.
Was there a snapshot feature, then we could have at least rolled back to the previous version and backup the zone files before rebuilding it.
SoftDux,
"Was there a snapshot feature, then we could have at least rolled back to the previous version and backup the zone files before rebuilding it."
No offence, but the same would have been achieved by using normal cloudmin backups. :-)
Yes, but with cloudmin backup the machines goes down for a while
I would also like a snapshot functionality, thanks in advantage!
Lucian,
Please don't insult me (http://www.urbandictionary.com/define.php?term=no%20offence) unless you actually know what a snapshot is, how it works and how much quicker and easier it is to restore than "cloudmin backups".
Have you ever had to take down a 200GB Virtual machine full of live website todo a backup or restore? And how did your clients feel about it?
Getting back to the original question though. VMWare, Virtuzzo, Citrix, OnApp, etc all offer instant snapshots which only backup what has changed since the last snapshot and can also restore back to that snapshot in an instance. CloudMin's current backups are pretty much useless, slow and redundant
I really didn't mean to insult anyone. Really! :)
Now, yes, cloudmin's backups may be not that great, but I wouldn't call them useless and redundant.
Thing is Jamie & Joe can only do so much now and it's a great effort anyhow, imagine that they not only need to know very good Perl, but also the internal workings of linux memory management, kvm, xen, lvm, file disks, mysql, apache, postgresql, proftpd et caetera.
Personally I think that we are just asking too much of them right now.
If I were you and my business depended on this feature I'd contact Jamie and propose a sponsorship for this feature. Think about it. How much would this be worth for you? €500, €1000?
My 2 pence.
Lucian, WTF??? Are you serious???
"Personally I think that we are just asking too much of them right now."
Great applications don't become great cause the developers make it great, but because they incorporate features that their clients request.
For a 1000Euros I can buy VMWare, then why would I even bother with CloudMin?
Actually, the single-host version of VMWare ESXi is free.
That aside:
Great applications don't become great cause the developers make it great, but because they incorporate features that their clients request.
Uhm... Say what? While client requests sure play quite a role, I don't think it's adequate for you to belittle the effort that the three-people (!!) developer team (of which only one or two are actually doing the coding, as far as I understand) is making here!
heh, I knew what I said could be seen from the wrong point of view.
The CloudMin team has done a great job so far. But, and the point I'm trying to make, CloudMin would probably have stayd a hobby / in-house project if they never took feature requests from their clients seriously.
Differently put, why would you ever buy an application if it doesn't do what you need it to tod? Any great appliation has a lot of end-user requested features.
+1