I would be happy with that too - my concern is the problem creeping in again. If you search the net - you will find that this has been an ongoing issue in the redhat kernel for a very long time.I suggest to debug LVM instead (Besides, LVM works for many,many people!)
it's true, but it would be better to support hcp in vzdump, at least until the lvm issue will be fixed (who knows when), as an alternative to the minority who have the lvm issue.I suggest to debug LVM instead (Besides, LVM works for many,many people!)
Simply claiming that something is better is easy. But I would prefer hard numbers - so please provide a benchmark or test case that shows that HCP is better (or reproduces the LVM crash). We can then try to debug an make things better.Try it! It is rather good
I can reproduce the lvm crash on my systems easily. Simply generate high io... Do a 'very long' backup. And the unmount the snapshot.Simply claiming that something is better is easy. But I would prefer hard numbers - so please provide a benchmark or test case that shows that HCP is better (or reproduces the LVM crash). We can then try to debug an make things better.
As far as I followed your issue most are related to lsi raid cards with the latest 2.0 kernel - is this true?Maybe I should log one then .
To date I have been using these forums to post the issue - guess that is the wrong location.
Overall - lvm is stable. However... I have 17 servers all different specs.
6 of them crash. The rest don't.
I am certain that this is related to disk io / driver issues. But to date; no 100% fix has been found. As a result - i have looked at hcp to solve the problem. This works perfectly
No hangs, lookups or general crashes when backing up.
we don´t think that its a good idea to include non-free software in a free software product, this makes no sense for me. btw, instead of lvm snapshot there will be also qemu live copy or even btrfs sound promising.I am happy to do the coding? Really quite a simple job to include the function.
Sent from my iPhone using Tapatalk
from hcp homepage:I also don't see how that can work for block devices we pass directly to KVM. IMHO that will not work or is at least very dangerous.
no, you are just wrong. we suggested that you help to fix the issue in LVM. additonally I told you two alternative solutions for the future.I totally get the point you guys may have in keeping to lvm - and reluctance to use alternative systems.
Remember.... This is all about choice. No solution is a perfect fit. We all do things a bit differently.
The beauty of open-source is that it provides the ability for alternative ideas and solutions (work a rounds) to be presented and used.
Sent from my iPhone using Tapatalk
We simply state our opinion, and there is no negative attitude here (sorry if you feel that way - that was not our intention).Why oh why does written text get miss-interpretted!
I feel you have taken my comments as negative or critisisms - yet that is not what I have intended.
You did not provide any details, and there is no reproducible test case - so it is really hard to work on a fix.- I have raised the point that there is an issue with LVM snapshots not working well across all systems. From my point this is a big issue as it kills off a very effecting backup system. (I normally do vzdumps weekly.. and bacula daily)
No, because we do not want to support that. Instead, we want our users to use open-source (LVM).- is is possible to get an option added to vzdump to allow hcp backups as an alternative to lvm?