On Mon, Sep 20, 2010 at 05:37:46PM +0200, Roman Vesely wrote: > I am a sponsor and satisfied user of an excellent project > Linux-Vserver. thanks, and good to hear! > Nowadays I have been using distributive Vserver for Debian > Etch and Lenny on more than 20 servers without any bigger > problems. means the distro kernels/tools? > I would like to start testing the last experimental version, > but situation relates to current state of the project is > very confusing and unnoticed for people who did not watch > mailling list or irc in detail. what confuses you? > I prefer kernel for Debian. My questions are: again, you mean distro kernels, yes? > * What are the changes between version 2.3.0.36.27 which is > in a current Squeezy and version 2.3.0.36.29.6? I presume that you are again referring to the debian kernel (2.3.0.36.27), so the problem there is, I do not know what debian patched in addition to 2.3.0.36.27, but I can give you a list of changes if you provide that patch > Can I use distributive Squeezy version or are there any > serious mistakes? we know that the 'stable' debian kernels are horribly broken but most of the testing kernels should probably be fine, give or take a feature/bugfix > * What are the changes between 2.3.0.36.29.x a 2.3.0.36.30.x ? if you are referring to the patches against mainline, then the changes are mostly in the way Linux-VServer integrates with the mainline changes (regarding virtualization) > * What are the changes between 2.3.0.36.30.x a 2.3.0.36.32.x? same here, if you need precise details, just ask, there is no official changelog for those > * Is the table http://linux-vserver.org/Feature_Matrix which > is 3 years old up-to-date? probably not, but note that Linux-VServer is a quite mature project, so the big number of changes happened like 8 years ago and it is fairly feature complete since several years now > * In which contidition or state is the implementation > HardCPULimit in new experimental versions? depends on the actual patch, before mainline introduced the CFS scheduler (completely fair scheduler), we used the TBE (token bucket extension) to handle fair distribution and hard limits for guests, with the introduction of the quite promising CFS scheduler, we gradually removed the TBE and adopted the new design ... some versions incorporate an experimental patch to allow for hard limits, but mainline is working (or whatever they call it :) on a proper hard limit implementation for almost a year now and the patches are incomplete and outdated, so we do not include them in recent versions > I intercede for a continous publishing of changes in the new > releases. we do so for stable releases, but the experimental branch is still in flux and maintaining accurate changelogs would consume quite some time, but feel free to hang around on the irc channel, ask about changes and document them as some kind of changelog (which I'll gladly review and comment on) > * Do you plane a stable release from experimental branche > in a near future? we have been planning to get there for quite some time now, but the mainline virtualization effords cause quite some fluctuation in the code areas we patch and so the kernel itself becomes a moving target ... stabilization involves testing and code review and there is not much point in reviewing code which has to be changed completely the next week > * Will it be supported Debian Squeezy kernel? no idea, probably a question for the debian folks > If not it would be possible to make financial support? I think it is more a political question than a financial but as financial support is always welcome, that might help as well :) > Probably there are more people who would be interested in it. I guess so, at least the percentage of debian users on the Linux-VServer channel is quite hight > Thank you a lot for your answers, I appreciate it, > you save my time. you're welcome! HTC, Herbert > Roman