Home » General » Support » There is great concern that OpenVZ is no longer being supported.
There is great concern that OpenVZ is no longer being supported. [message #38005] |
Sun, 08 November 2009 19:34 |
|
JimL
Messages: 116 Registered: February 2007
|
Senior Member |
|
|
See the discussion on the General/discussions forum page.
We have asked over there about the status of newer kernels but no one from the development team has responded. A number of us are about to jump ship and start deploying different solutions for implementing virtual systems thinking that OpenVZ may be dead or dying. I don't want to but it looks like OpenVZ is no longer being maintained. What say the developers?
Jim.
[Updated on: Sun, 08 November 2009 19:35] Report message to a moderator
|
|
|
|
|
Re: There is great concern that OpenVZ is no longer being supported. [message #38021 is a reply to message #38020] |
Mon, 09 November 2009 12:32 |
|
All the kernel updates are published at announce@ mailing list (http://wiki.openvz.org/Mailing_lists) AND at http://wiki.openvz.org/News/updates (also see the sidebar at the main wiki page). The last update was released last Saturday, fixing a serious security flaw. Yes the kernel support team and the QA team had to work last Saturday until about 11pm in order to release this update, because of its importance.
The only stable and maintained branches we have now are RHEL4 and RHEL5-based. For those branches updates are guaranteed (if we can talk about guarantees here).
2.6.24, 2.6.26 and 2.6.27 are all development branches, never ever recommended for production, and yes, we kinda stopped maintaining those.
We will probably be opening another devel branch soon, based on 2.6.32.
Speaking of tools, take a look at activity in vzctl's git, you'll see quite a number of new patches flowing it.
Ergo, OpenVZ is not dead. That should not preclude you from trying out other solutions if you feel like it. Diversity is good.
Kir Kolyshkin
|
|
|
|
|
|
|
|
Re: There is great concern that OpenVZ is no longer being supported. [message #38038 is a reply to message #38037] |
Tue, 10 November 2009 23:19 |
|
Ales wrote on Wed, 11 November 2009 01:49 | Is the difference between the redhat's 2.6.18-128.x and 2.6.18-164.x kernel so great, that it's easier for your team to backport fixes into an older base?
|
I have already answered this elsewhere (http://bugzilla.openvz.org/1358) so below is semi copy-paste.
The last update was a high priority security update, and for such updates we prefer to not do a full rebase, but rather take the proven stable already well-tested kernel and just patch the security hole(s) in order to minimize potential stability problems and make a QA cycle faster. This is exactly what we did this time. The CVE is dated Nov 3, our updates were published Nov 7 -- we just can not achieve it with a full QA cycle.
Quote: | I assumed that because of Redhat's consistency of the kernel ABI/API, moving on should be pretty straightforward. I guess it isn't?
|
Now it's not -- internal kernel changes are quite significant and we have to port our code.
Having said that, we already have some test builds based on that kernel, they are in QA and will eventually be released. It is indeed a lot of changes and therefore requires a lot of testing (i.e. a full QA cycle). During the QA we usually find bugs in both our code and Red Hat's code (I blogged about it about a year ago), so it's definitely worth it.
Kir Kolyshkin
|
|
|
Re: There is great concern that OpenVZ is no longer being supported. [message #38040 is a reply to message #38005] |
Wed, 11 November 2009 02:45 |
mhw
Messages: 12 Registered: March 2007
|
Junior Member |
|
|
Some "features" do not get backported for a variety of reasons. One amongst several for me right now is the MD5 signatures on TCP sessions (CONFIG_TCP_MD5SIG) which is required for some BGP sessions when peering with ISP's that require it. It's not in 2.6.18 nor would I expect it to ever get backported (limited applicability and non-trivial). There's also some IPv6 stuff that is significantly improved in more recent kernels for standards compliance.
The question arises, though, with 2.6.32... How is the support going to manifest itself? Will the vz utilities be enhanced to utilize the lxc containers that everyone has been working so hard on or is it going to be another kernel patch for the vz feature set? Is there something missing from the lxc implementation / feature set which infeasible?
[Updated on: Wed, 11 November 2009 02:46] Report message to a moderator
|
|
|
Re: There is great concern that OpenVZ is no longer being supported. [message #38121 is a reply to message #38021] |
Thu, 19 November 2009 18:46 |
pva0xd
Messages: 24 Registered: February 2008
|
Junior Member |
|
|
kir wrote on Mon, 09 November 2009 15:32 | 2.6.24, 2.6.26 and 2.6.27 are all development branches, never ever recommended for production, and yes, we kinda stopped maintaining those.
|
This basically means that you stopped support Gentoo.
We are unable to use rhel based kernels: they are too old to be used with newer toolchain and there is really large number of configurations that fail to build.
So currently I'm not sure what to do next: should we keep openvz kernel in Gentoo or should we suggest users other alternatives? 2.6.27 openvz kernel is just security nightmare as at least two really important security issues are there. Fortunately 2.6.27 is long maintained kernel upstream so I've managed to backport fixes from stable branch into openvz sources. But there are other openvz related bugs and even some fixes still sit in bugzilla! I really don't understand why it's so hard to add fixes straight to git and since I'm not capable to read all reports at openvz's bugzilla and merge fixes from there into our patchset I really wonder: is it worth to keep openvz-sources in Gentoo? For me it looks like it's not very sane to encourage users use unmaintained technology...
And yes, there is sense to use Gentoo on HN since it makes possible to build everything with hardened toolchain and really drop everything unneeded.
[Updated on: Thu, 19 November 2009 18:47] Report message to a moderator
|
|
|
|
|
|
|
|
|
Re: There is great concern that OpenVZ is no longer being supported. [message #38385 is a reply to message #38383] |
Sun, 13 December 2009 14:35 |
Paparaciz
Messages: 302 Registered: August 2009
|
Senior Member |
|
|
aTan wrote on Sun, 13 December 2009 15:39 | I'm aware of it, but 2.6.18 is just old. And as it was said above doesn't fulfill todays needs even with some backported stuff. Could you just say you won't support newer kernels in the near future (maybe until RH customers will need it), so people can stop hoping for it and try find some other solutions for their needs or at least put up with this? I guess people will appreciate it and also won't bother you with this question in the future. Sticking it in Support forum will be also fine.
|
what exactly doesn't work for you in rhel kernel 2.6.18?
also keep in mind that in next year redhat should release rhel6 version which will be based on newer kernel (i guess that it would be 2.6.3x).
|
|
|
|
|
Re: There is great concern that OpenVZ is no longer being supported. [message #38389 is a reply to message #38388] |
Sun, 13 December 2009 15:59 |
Paparaciz
Messages: 302 Registered: August 2009
|
Senior Member |
|
|
pva0xd wrote on Sun, 13 December 2009 17:49 | kir wrote on Sun, 13 December 2009 14:36 |
Again (and again): you do not have to use RHEL in order to use RHEL-based kernel. Believe it or not, you can run this kernel on most of the other distros just fine.
|
Please, stop telling this nonsense. No offense, but I just did short search on what kernels use other distributions:
altlinux: 5.0 (current) branch uses 2.6.27 kernel
suse (11.1): 2.6.27
mandriva: 2.6.26
debian(!): lenny uses 2.6.26
|
in all modern distributions you should not consider version number as it is.
redhat backports a lot of stuff to that kernel from more modern kernels. other distributions does the same for all packages not only kernel.
|
|
|
|
Re: There is great concern that OpenVZ is no longer being supported. [message #38391 is a reply to message #38385] |
Sun, 13 December 2009 16:31 |
|
aTan
Messages: 29 Registered: March 2008
|
Junior Member |
|
|
Paparaciz wrote on Sun, 13 December 2009 15:35 | aTan wrote on Sun, 13 December 2009 15:39 | I'm aware of it, but 2.6.18 is just old. And as it was said above doesn't fulfill todays needs even with some backported stuff. Could you just say you won't support newer kernels in the near future (maybe until RH customers will need it), so people can stop hoping for it and try find some other solutions for their needs or at least put up with this? I guess people will appreciate it and also won't bother you with this question in the future. Sticking it in Support forum will be also fine.
|
what exactly doesn't work for you in rhel kernel 2.6.18?
also keep in mind that in next year redhat should release rhel6 version which will be based on newer kernel (i guess that it would be 2.6.3x).
|
For me (small company's needs) it's just: no xtables, no upstream IPv6, some 3rd party patches (imq, (e)sfq, etc) are provided only for new kernels, it's slower (not such a big deal). And I'm sure there will be a lot more issues if I'll try to downgrade from 2.6.27 (udev needs at leas 2.6.25 since version 145, newer toolchain (gcc4.4, binutils 2.20) can has problems with 2.6.18, etc). And I'm sure there must be a lot other problems I'm not aware of. It's not only about last number in a kernel version.
|
|
|
Re: There is great concern that OpenVZ is no longer being supported. [message #38392 is a reply to message #38391] |
Sun, 13 December 2009 17:08 |
Paparaciz
Messages: 302 Registered: August 2009
|
Senior Member |
|
|
aTan wrote on Sun, 13 December 2009 18:31 |
For me (small company's needs) it's just: no xtables, no upstream IPv6, some 3rd party patches (imq, (e)sfq, etc) are provided only for new kernels, it's slower (not such a big deal). And I'm sure there will be a lot more issues if I'll try to downgrade from 2.6.27 (udev needs at leas 2.6.25 since version 145, newer toolchain (gcc4.4, binutils 2.20) can has problems with 2.6.18, etc). And I'm sure there must be a lot other problems I'm not aware of. It's not only about last number in a kernel version.
|
ok, if this features so mission critical, why you don't try development openvz 2.6.27,2.6.26 versions?
I think that people miss main point- openvz dev team does what it can at best. rhel kernel is enterprise level, a lot of stuff is backported, but also there are not so much changes in minor releases, as it stands for enterprise, what means better older, but stable. Also think what does it means to keep in touch with newest kernel.org releases. you simple have to give too much time, because while you are looking at and adopting to some kernel release, at this time again could be next release.
I don't understand animosity from people. Look at openvz project as community driven. you as not happy user of project can give it back by adopting openvz to newest kernels. You even can hapily fork whole project. But just pointing that you(openvz team) are not doing what i want is silly.
I think that openvz made best decision to keep with rhel kernel. many hosting and virtualization service providers uses rhel based distributions (for example centos), and are very happy with that. what does it mean to choose bleeding edge kernel for hosting company? it means more often patching, rebooting and from that derives worse service availability.
For me, and for company where I work, stability is more important than super extra new features.
|
|
|
|
Re: There is great concern that OpenVZ is no longer being supported. [message #38394 is a reply to message #38390] |
Sun, 13 December 2009 17:50 |
pva0xd
Messages: 24 Registered: February 2008
|
Junior Member |
|
|
kir wrote on Sun, 13 December 2009 19:28 |
What I said (and you can clearly see this from my quote above)
(1) is you do not have to use RHEL in order to use RHEL-based OpenVZ kernels;
(2) You can run RHEL5 based OpenVZ kernel on most of the other (i.e. non-RHEL5) distros.
Now, you are telling me it's nonsense and provide some information about what kernels some distros are using.
I am sorry but I can not get the connection here -- why the information that ALTLINUX uses 2.6.27-based kernel (and Mandriva uses 2.6.26-based one, and so on) lead you to the conclusion that I'm telling nonsense (i.e. either my point (1) or (2) above (or both) is/are invalid)?
|
Ok, let me explain this. I showed you that distributions do not use RHEL kernels. Since there is zero probability that anyone manages to dig openvz related fixes from one big hunk of code you provide us with (RHEL5 patchset + openvz patchset in one go), I'm sure that very few recent openvz related fixes get to distributions. This means that openvz upstream does not support all distributions but RHEL based and since the topic of this discussion is "There is great concern that OpenVZ is no longer being supported." repetition that 2.6.18 based kernels are supported is nonsense or at least does not justify that openvz is supported anywhere beside RHEL5.
kir wrote on Sun, 13 December 2009 19:28 | Believe me or not, we are doing the best we can -- for now we only have resources to support RHEL5-based kernels, plus I hope we will start porting to 2.6.32/RHEL6 in some non-too-distant future.
|
While new branch sounds great, without change in workflow it just makes problem a bit less visible, but still new branch will not fix the problem. The greatest concern is that at some point of time after creation of new branch it will stop to evolve and while we will see new RHEL based kernels and ChangeLog will tell us that some openvz related fixes are there, git will keep silence and these fixes became inaccessible for those who are unable to use RHEL kernels. The only change in workflow I'm looking for is - please, update git whenever something was fixed in openvz. It is understandable why you use RHEL patchset after all and why production kernels are based on it, but it's hard to see why patches are provided as one big hunk of code without real possibility to separate RHEL changes from openvz. So, please, if you start new branch do not abandon it, like it is now! Just commit fixes there so we could see, test and use the progress. As a side result of this workflow you will have more beta testers and this will help you to release more well tested kernels (then with internal testing process you have now). This workflow looks easy but probably I'm completely wrong here and there are reasons to drop git at some point of time and leave community with known bugs.
Any way thanks, that you still answer here, Kir. It's really appreciated that upstream at least tries to hear us.
|
|
|
Re: There is great concern that OpenVZ is no longer being supported. [message #38395 is a reply to message #38390] |
Sun, 13 December 2009 22:12 |
Lorddusty
Messages: 6 Registered: January 2008 Location: Cologne, Germany
|
Junior Member |
|
|
Hi,
well, it's nice to read (although it's not new to me) that RHEL Kernel would work on other distros... I mean, who is wondering, we are all talking about _LINUX_ distros.
Anyways, as already mentioned earlier, lot's of userspace tools, e.g. iptables or apr require special features in Kernel. Even if those features are being backported into 2.6.18 RHEL, the user-space configure does not know about it and as it uses uname to figure out, if the right kernel-version is being used, those userspace tools simply do not compile.
Another example is that OpenSuSE 11.2, which is the current stable-release, is not running without dirty-hacks (means recompilation of source-rpms from 11.1) as neither apache, nor postfix-rpms from this distros are working with neither latest RHEL 2.6.18 nor 2.6.24 (didn't try 2.6.27 branch as there DRBD was not running properly last time I tried)
Also the bug with IPv6-PMTU ... I really still would like to read an answer. I mean, I spent couple of hours to trace down where this came from, and I fed not to few info into bugzilla, but supporting a project by giving qualified bug reports seems not to make sense if there comes no reaction for several months.
I really like this project, and I really appreciate that at least KIR is replying to this thread, but I do not really see a perspective if both communication and development goes on in the current way.
Best regards
Jens
|
|
|
|
|
|
Re: There is great concern that OpenVZ is no longer being supported. [message #42729 is a reply to message #42720] |
Mon, 16 May 2011 14:34 |
drobbins
Messages: 1 Registered: May 2011 Location: Albquerque, NM
|
Junior Member |
|
|
pva0xd wrote on Sun, 15 May 2011 04:11Paparaciz, please read topic before answer. I've already told why rHELL kernels are bad idea for Gentoo. Although may be I'll create ebuilds for them I'll doubt I'll be use/test them.
I am the creator of Gentoo Linux and currently develop Funtoo Linux (docs.funtoo.org). I like OpenVZ and we maintain compatibility with OpenVZ RHEL5 kernels. We have RHEL5 stable and RHEL6-testing kernels available. I actually just committed new ebuilds for both. You will need true Funtoo Linux install (funtoo unstable/current recommended) to use them, though, because we have a different udev than Gentoo.
sys-kernel/rhel5-openvz-sources has the latest RHEL5 sources. Use with udev-146-r3 and it works. There are (not yet 100% updated) instructions on our wiki on how to do the downgrade easily. - Also, this ebuild automatically installs gcc-4.1.2 and uses it to build, so it is using the recommended GCC version for this kernel.
sys-kernel/rhel6-openvz-sources has the lastest RHEL6-testing sources.
Both ebuilds can be built with the "binary" USE flag which will ALSO install a standard binary kernel built using genkernel with the recommended OpenVZ kernel config.
I have also added several fixes and improvements (Funtoo compatibility) to our version of vzctl.
If you are an OpenVZ fan, I'd welcome you to try Funtoo Linux and give me your feedback because OpenVZ support in Funtoo Linux is a personal priority of mine.
Regards,
Daniel
|
|
|
|
|
|
|
|
Re: There is great concern that OpenVZ is no longer being supported. [message #42984 is a reply to message #38005] |
Fri, 24 June 2011 16:54 |
mustardman
Messages: 91 Registered: October 2009
|
Member |
|
|
I am very very impressed with OpenVZ. Very stable and scales incredibly well!
The best VM solution out there for what it does IMHO. For what I am doing nothing else even comes close. The big one for me is memory usage which is typically at a premium. On a solution like Xen the memory is reserved. So whatever the VM's have assigned to them is no longer available even if they are not using some of it.
On OpenVZ you always have access to all available memory. It doesn't sound like a big deal until you see it in action. When you have 20 VPS's on a server and each one has maybe 100MB of free memory that is 2Gig of extra memory you have to play with. Also keep in mind this is averaged over a bunch of different users. So a spike in usage at different times by different users is not going to change that availability. That means you can add a lot more VPS's for a given amount of memory than you could with other solutions.
And if you want to oversell (which I wouldn't recommend for any solution) OpenVZ just starts using the swap file.
CPU sharing appears to be very efficient as well. What I am doing involves real time applications and nobody ever complains about any jitter or delay. As long as I keep total average CPU usage below about 50% it's not a problem. And with OpenVZ's ability to divide up CPU cycles I simply set it so that no one VPS can hijack all the CPU cycles.
Anyways, I sure hope OpenVZ is able to keep a base of support. It would be a shame if it went away and forced people to use inferior solutions. I think as long as Parallels is making money OpenVZ will be ok. As far as I can tell Parallels is still committed to OpenVZ as their open source proving ground.
[Updated on: Fri, 24 June 2011 17:12] Report message to a moderator
|
|
|
|
Goto Forum:
Current Time: Mon Dec 30 17:12:49 GMT 2024
Total time taken to generate the page: 0.05729 seconds
|