Re: Earth Computing
Jordan Erickson wrote: > Etienne Goyer wrote: > *snip* >>> Is LTSP not heavily dependant on NFS? >> Yes, as the read-only root file system. Security is not critical in >> this use-case. > > Actually, since Ubuntu 8.04 (or earlier?), The LTSP client filesystem > uses NBD by default. > > http://en.wikipedia.org/wiki/Network_block_device My bad, you are absolutely right. I still live in 2006, sorry! :) -- Etienne Goyer Technical Account Manager - Canonical Ltd Ubuntu Certified Instructor -LPIC-3 ~= Ubuntu: Linux for Human Beings =~ -- ubuntu-server mailing list ubuntu-server@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-server More info: https://wiki.ubuntu.com/ServerTeam
Re: Earth Computing
Etienne Goyer wrote: *snip* > >> Is LTSP not heavily dependant on NFS? > > Yes, as the read-only root file system. Security is not critical in > this use-case. Actually, since Ubuntu 8.04 (or earlier?), The LTSP client filesystem uses NBD by default. http://en.wikipedia.org/wiki/Network_block_device -- Jordan Erickson http://www.logicalnetworking.net * Please consider the environment before printing this e-mail * -- ubuntu-server mailing list ubuntu-server@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-server More info: https://wiki.ubuntu.com/ServerTeam
Re: Earth Computing
On Thu, 28 Jan 2010, Etienne Goyer wrote: > Alvin wrote: > > I know. It's still the most useful filesystem around, and we want to > > use the best tools for the job. LVM is not as flexible as ZFS. > > The main objection I have against ZFS is that it does not have any > recovery tool (or so I heard; correct me if I am wrong). As such, I > would not trust it in production, but you may know something I don't. > > > These run Jaunty because of the above bugs and because of a regression > [bug > > 224138] "No NFS modules in karmic 32-bit" > >> Again, not trying to make excuse, and not sure I understand the problem > >> correctly, but that sounds like an overstatement. It seems like the > >> -virtual kernel flavor is missing some modules (including those for > >> NFS*v4*), but you could just as well use the -generic or -server flavor. > >> Or am I misunderstanding something? > > > > Yes, virtio. The virtual kernels perform faster. In extreme cases, a > > calculation could take as much as 6 hours. You will run in stability issues > > when using the normal kernel. (Our old Solaris8 machines do the same in 6 > > days.) > > Besides, when the virtual machines were build, we had no choice due to this > > bug: [kvm guests not using virtio for networking lose network connectivity] > > https://bugs.launchpad.net/bugs/286101 > > I do not know. I have the 2.6.31-17-generic kernel running on karmic, > and it does have virtio compiled as a module. I run all my KVM vm using > either the -generic or -server kernel flavour, and I have not had any > problem so far. Are you sure you *have* to use the -virtual kernel? I'm not completely following the above, maybe you're saying you chose -generic/-server in the hardy time frame (the bug is from hardy). In karmic and lucid, the -virtual kernel is a literal subset of -generic-pae on i386 -server on amd64 If you compare checksums they're identical. The kernel team calls it a "sub-flavour". This should absolutely not cause any differences in performance. If you see differences, its not related to this. In karmic and lucid for sure, the only benefit of -virtual is disk footprint in the guest. > Someone else can clarify, but I was under the impression the -virtual > flavour was being phased out. Is that correct? -virtual is not going away as far as I am aware. In fact, the most recent discussion on -devel was that it might someday need to be a separate entity ("flavour") than a strict subset of -generic-pae/-server ("sub flavour"). That, though, will only happen if there is good reason for certain drivers to be compiled into -virtual that might cause problems in -server/-generic-pae. -- ubuntu-server mailing list ubuntu-server@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-server More info: https://wiki.ubuntu.com/ServerTeam
Re: Earth Computing
Alvin wrote: > I know. It's still the most useful filesystem around, and we want to > use the best tools for the job. LVM is not as flexible as ZFS. The main objection I have against ZFS is that it does not have any recovery tool (or so I heard; correct me if I am wrong). As such, I would not trust it in production, but you may know something I don't. These run Jaunty because of the above bugs and because of a regression [bug 224138] "No NFS modules in karmic 32-bit" >> Again, not trying to make excuse, and not sure I understand the problem >> correctly, but that sounds like an overstatement. It seems like the >> -virtual kernel flavor is missing some modules (including those for >> NFS*v4*), but you could just as well use the -generic or -server flavor. >> Or am I misunderstanding something? > > Yes, virtio. The virtual kernels perform faster. In extreme cases, a > calculation could take as much as 6 hours. You will run in stability issues > when using the normal kernel. (Our old Solaris8 machines do the same in 6 > days.) > Besides, when the virtual machines were build, we had no choice due to this > bug: [kvm guests not using virtio for networking lose network connectivity] > https://bugs.launchpad.net/bugs/286101 I do not know. I have the 2.6.31-17-generic kernel running on karmic, and it does have virtio compiled as a module. I run all my KVM vm using either the -generic or -server kernel flavour, and I have not had any problem so far. Are you sure you *have* to use the -virtual kernel? Someone else can clarify, but I was under the impression the -virtual flavour was being phased out. Is that correct? > Is LTSP not heavily dependant on NFS? Yes, as the read-only root file system. Security is not critical in this use-case. > I think it's a mistake to throw the Unix > way overboard in favour of MS Windows solutions. NFS and CIFS have different > usage scenarios. CIFS has been an IETF standard draft for a long time (now expired). It is not Windows-specific anymore (and has not been for a while). With the advent of the Unix Extensions, it offer complete POSIX file system semantic. What more can you ask for? :) > There is certainly room for improvement, but that's why NFSv4 > exists. I am not experienced enough in NFSv4 to make an enlightened judgment, but my feeling is that it is quite byzantine. Beside the dependence on Kerberos for authentication (which is perfectly fine, but rule it out in many use-case), it still relies on a mish-mash of different RPC mechanism for id mapping, locking, etc. I could never wrap my head around it. That being said, Samba is anything but straightforward itself, and I am not apologizing for it. > Users need access to the same filesystems anyway. Of course they do make > mistakes and erase files, or move them to the wrong place. That's why the > regular ZFS snapshots can be so handy. Just so you know, LVM can also do CoW snapshot (albeit, perhaps, less elegantly than ZFS). Best regards, -- Etienne Goyer Technical Account Manager - Canonical Ltd Ubuntu Certified Instructor -LPIC-3 ~= Ubuntu: Linux for Human Beings =~ -- ubuntu-server mailing list ubuntu-server@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-server More info: https://wiki.ubuntu.com/ServerTeam
Re: Earth Computing
On 28.01.2010 14:01, Alvin wrote: > - The product has been proven to be reliable > - Searches are a lot faster (last time I checked). Well, I must admit that netatalk is one of most reliable services I have ever seen on UNIX. To be honest, I find it strange that someone is using anything else. > - CIFS is not an option for Mac in some environments. Have you tried it? I'd agree with this statement. CIFS/SMB implementation on MACOS, even though it is based on samba, is very bad. > On the whole, Macs do not play well in a networking environment. Adobe on Mac > certainly does not make things easier. That's true. Saving files to network device from Adobe software on Mac is lottery. As for apple, at some stage, MacOS couldn't print on other CUPS print servers. Apple changed CUPS source to make it not work with anything else than other MacOSes. Luckly, that was fixed and Apple bought CUPS. -- ubuntu-server mailing list ubuntu-server@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-server More info: https://wiki.ubuntu.com/ServerTeam
Re: Earth Computing
On Wednesday 27 January 2010 20:17:00 Joe McDonagh wrote: > Alvin wrote: > > A lot of questions in the annual user survey concern cloud computing. I > > administer some small businesses and use Ubuntu in most of them. Maybe my > > biggest client will one day use a personal cloud, and I applaud the > > efforts, but I can 't help but notice that other things are left in the > > cold. [...] > IDK what's with the 'hey brah' subject line, but you probably shouldn't > be using an LTS-interim release like karmic or jaunty for the business. > You will run into more show-stopping bugs. I prefer RHEL for business, > but whatever. This is true in a lot of situations, but Hardy is not an option for us. We're using the virtual servers to do (a lot) of assembling of .xml files. That's a lot of integer calculating and pretty heavy I/O over NFS. I tested Hardy, Intrepid, Jaunty and Karmic. Karmic has the only version of KVM that does not crash while doing those calculations. RHEL is actually the supported Linux distribution for the software we're using (http://www.sdlxysoft.com/en/products/xpp/), and we seriously considered it. The main reason we tried Ubuntu, is because I know the Debian way better and that's an important consideration. The problems with LD_LIBRARY_PATH and libmotif do not exist in RHEL, but we found workarounds for them. > And for the firewall you might want to think about moving to OpenBSD. OpenBSD is certainly made for that job, but the out-of-the-box solution we're using now is doing great for now. (http://www.collax.com/) Besides, they have really excellent support. -- ubuntu-server mailing list ubuntu-server@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-server More info: https://wiki.ubuntu.com/ServerTeam
Re: Earth Computing
On Wednesday 27 January 2010 23:15:37 Dustin Kirkland wrote: > On Tue, Dec 15, 2009 at 6:40 AM, Alvin wrote: > ... > > > This is a real-life scenario. Is it common? I don't know. It's not free > > of struggles as you can see. So, this is a plea for quality. Cloud > > Computing might be very important, but please don't lose sight of the > > little guys who just want some 'classic' servers. > > As one of the Canonical Ubuntu Server Developers working on our Ubuntu > Enterprise Cloud, I'll respond to this ... > > We absolutely have a focus on quality for the "classic servers", as > you call them. Basic Ubuntu Servers are our foundation, truly, for > all of our Cloud Computing efforts, both as a Virtualization Guest, > and a Virtualization Hypervisor. > > Your email is well constructed, and having a targeted bug list is an > excellent start. My guidance to you would be: > * Help get each of those bugs into a triaged state, and with an > appropriate priority set, and filed against the correct package. > * Re-test them with the latest development code, the 10.04 release > that's under development. > * Provide all the information you possibly can to help a developer > work to solve those problem. > * Nominate them for release, probably targeting Ubuntu 10.04 Lucid at > this point. > * If you have hacks, work arounds, fixes, or patches, please, by all > means, attach them to the bugs, as again, this will help a developer > get an appropriate fix into Ubuntu sooner than later. > * And have just a bit of patience and understanding that we have 100s > (or 1000s) of bugs, and merely a 5-person Ubuntu Server Team in > Canonical (plus a few dozen more Ubuntu and Debian community members > that help out). Thanks. I appreciate the suggestions and all your work on this. It's about time I start testing Lucid. Eventually we will make the switch and then stay on Lucid on all non-desktop machines until the next LTS release. -- ubuntu-server mailing list ubuntu-server@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-server More info: https://wiki.ubuntu.com/ServerTeam
Re: Earth Computing
On Thursday 28 January 2010 09:31:00 Etienne Goyer wrote: > > Alvin wrote: > >> Why not Ubuntu? > >> - ZFS (does not need much explanation) > > Not looking to make excuse, but just so you know, ZFS on Linux is > unlikely to happen due to licensing issue: > > http://en.wikipedia.org/wiki/ZFS#Linux I know. It's still the most useful filesystem around, and we want to use the best tools for the job. LVM is not as flexible as ZFS. Maybe things will brighten for Linux when BTRFS comes around. > >> These run Jaunty because of the above bugs and because of a regression > >> [bug > >> > >> 224138] "No NFS modules in karmic 32-bit" > > Again, not trying to make excuse, and not sure I understand the problem > correctly, but that sounds like an overstatement. It seems like the > -virtual kernel flavor is missing some modules (including those for > NFS*v4*), but you could just as well use the -generic or -server flavor. > Or am I misunderstanding something? Yes, virtio. The virtual kernels perform faster. In extreme cases, a calculation could take as much as 6 hours. You will run in stability issues when using the normal kernel. (Our old Solaris8 machines do the same in 6 days.) Besides, when the virtual machines were build, we had no choice due to this bug: [kvm guests not using virtio for networking lose network connectivity] https://bugs.launchpad.net/bugs/286101 > >> - [bug 374907] "libmotif3 crashes" > > <...snip...> > > >> Sometimes you hear: "it's open source. Don't complain and fix it > >> yourself." That's partly true. I'm not a programmer, but I was able to > >> patch libmotif3 to solve the crashes. > >> The kind people in ubuntu-bugs also managed to convince me that I could > >> package the new version of openmotif myself and put it in Debian. Maybe > >> I'll learn how to do that, so that bug can at least be closed. I can > >> understand that there is not a lot of interest in this package, but we > >> need it and will probably need it for some time to come. > > IIRC, the Citrix ICA client depends on OpenMotif (not sure which > version), so that bug would be a biggie indeed if it breaks the ICA > Client. We have been using the ICA Client on hardy without any problem > so far, but I am putting that on my radar. Thanks for the heads up, I > will be looking into it. Thanks! There is also a [needs-packaging] bug report: https://bugs.launchpad.net/bugs/462182 > >> What I can't understand is that there would be no interest in NFS. Is > >> everyone using samba between unix machines these days? > > To be honest, yes. NFS is only really useful for read-only share, as > NFS < v4 does not have any form of authentication, where CIFS mount can > be authenticated. It is still not good enough, as the file operation > themselves are not encrypted (supposed to come in Samba any time now), > but it is a step in the right direction. NFSv4, because of its reliance > on Kerberos, is too hairy to set up in most case. > > In general, I try to avoid NFS whenever possible, except for trivial > things. CIFS with Unix Extensions has been serving me well so far. Is LTSP not heavily dependant on NFS? I think it's a mistake to throw the Unix way overboard in favour of MS Windows solutions. NFS and CIFS have different usage scenarios. There is certainly room for improvement, but that's why NFSv4 exists. I can't remember if there were questions about file sharing technology in the Ubuntu Server Survey, but you should try to put it in a survey. I'm curious about the result. Also, keep in mind that in some situations, security is not important. Security often comes at a price (complexity and speed). I admit, central user management is only partly functional in this company. Our current security system is based on the concept of 'User Private Groups'. That works fine over NFS (not v4). In order to use Akonadi on shared /home you do need NFSv4 due to locking issues. Implementation is tricky though. Users need access to the same filesystems anyway. Of course they do make mistakes and erase files, or move them to the wrong place. That's why the regular ZFS snapshots can be so handy. Our network has been based around NFS for at least over 12 years, and it was quite a shock to see that computers did no longer boot after the update to Karmic. > But thanks for your feedbacks, you are doing the right thing. I am not > working in the distro team (I am in Corporate Services), so I cannot do > anything to help with your bugs directly, but I think it is a very good > thing that we get this kind of feedback. I must say I'm quite happy with this attention. Your offer for commercial support will certainly be accepted in light of the fact that there is interest to know what people are actually doing (or trying to do) with Ubuntu. -- ubuntu-server mailing list ubuntu-server@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-server More info: https://wiki.ubuntu.com/ServerTeam
Re: Earth Computing
On Thursday 28 January 2010 11:58:31 Etienne Goyer wrote: > Ante Karamatić wrote: > > On 15.12.2009 13:40, Alvin wrote: > >>- Helios, A commercial application to provide file and print sharing > >>for > >> > >> Macintosh. > > > > Is there something wrong with netatalk? It's an open source application > > that provides file and print sharing. For OSX, AFP is deprecated anyway > > (and printing works much better with CUPS, which is owned by Apple). > > Indeed, modern MacOS actually speaks CIFS, so Samba is all you need. > netatalk is only really useful when speaking to ancient Mac, it is > somewhat obsolete these days. I told the samba server to deny access to macintosh due to a multitude of problems. MacOS does speak CIFS, and netatalk has probably evolved since I last tested it, but there are different reasons here to choose Helios: - The product has been proven to be reliable - Searches are a lot faster (last time I checked). - CIFS is not an option for Mac in some environments. Have you tried it? Resource forks will get lost and Mac users will see files that aren't there or will not see files that are there. This is reportedly fixed in MacOS 10.6, but we're not willing to replace all those expensive PPC machines and Adobe software yet. - We still have Mac OS9 machines in active use. (I know...) - Other than that, Helios adds some functionality like OPI, but we're not using that anymore. On the whole, Macs do not play well in a networking environment. Adobe on Mac certainly does not make things easier. They officially do not support networked computers. Helios makes life a bit easier for admins. -- ubuntu-server mailing list ubuntu-server@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-server More info: https://wiki.ubuntu.com/ServerTeam
Re: Earth Computing
Ante Karamatić wrote: > On 15.12.2009 13:40, Alvin wrote: > >>- Helios, A commercial application to provide file and print sharing for >> Macintosh. > > Is there something wrong with netatalk? It's an open source application > that provides file and print sharing. For OSX, AFP is deprecated anyway > (and printing works much better with CUPS, which is owned by Apple). Indeed, modern MacOS actually speaks CIFS, so Samba is all you need. netatalk is only really useful when speaking to ancient Mac, it is somewhat obsolete these days. -- Etienne Goyer Technical Account Manager - Canonical Ltd Ubuntu Certified Instructor -LPIC-3 ~= Ubuntu: Linux for Human Beings =~ -- ubuntu-server mailing list ubuntu-server@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-server More info: https://wiki.ubuntu.com/ServerTeam
Re: Earth Computing
On 15.12.2009 13:40, Alvin wrote: >- Helios, A commercial application to provide file and print sharing for > Macintosh. Is there something wrong with netatalk? It's an open source application that provides file and print sharing. For OSX, AFP is deprecated anyway (and printing works much better with CUPS, which is owned by Apple). -- ubuntu-server mailing list ubuntu-server@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-server More info: https://wiki.ubuntu.com/ServerTeam
Re: Earth Computing
> Alvin wrote: >> Why not Ubuntu? >> - ZFS (does not need much explanation) Not looking to make excuse, but just so you know, ZFS on Linux is unlikely to happen due to licensing issue: http://en.wikipedia.org/wiki/ZFS#Linux >> These run Jaunty because of the above bugs and because of a regression >> [bug >> 224138] "No NFS modules in karmic 32-bit" Again, not trying to make excuse, and not sure I understand the problem correctly, but that sounds like an overstatement. It seems like the -virtual kernel flavor is missing some modules (including those for NFS*v4*), but you could just as well use the -generic or -server flavor. Or am I misunderstanding something? >> - [bug 374907] "libmotif3 crashes" <...snip...> >> Sometimes you hear: "it's open source. Don't complain and fix it yourself." >> That's partly true. I'm not a programmer, but I was able to patch libmotif3 >> to >> solve the crashes. >> The kind people in ubuntu-bugs also managed to convince me that I could >> package the new version of openmotif myself and put it in Debian. Maybe I'll >> learn how to do that, so that bug can at least be closed. I can understand >> that there is not a lot of interest in this package, but we need it and will >> probably need it for some time to come. IIRC, the Citrix ICA client depends on OpenMotif (not sure which version), so that bug would be a biggie indeed if it breaks the ICA Client. We have been using the ICA Client on hardy without any problem so far, but I am putting that on my radar. Thanks for the heads up, I will be looking into it. >> What I can't understand is that there would be no interest in NFS. Is >> everyone >> using samba between unix machines these days? To be honest, yes. NFS is only really useful for read-only share, as NFS < v4 does not have any form of authentication, where CIFS mount can be authenticated. It is still not good enough, as the file operation themselves are not encrypted (supposed to come in Samba any time now), but it is a step in the right direction. NFSv4, because of its reliance on Kerberos, is too hairy to set up in most case. In general, I try to avoid NFS whenever possible, except for trivial things. CIFS with Unix Extensions has been serving me well so far. But thanks for your feedbacks, you are doing the right thing. I am not working in the distro team (I am in Corporate Services), so I cannot do anything to help with your bugs directly, but I think it is a very good thing that we get this kind of feedback. -- Etienne Goyer Technical Account Manager - Canonical Ltd Ubuntu Certified Instructor -LPIC-3 ~= Ubuntu: Linux for Human Beings =~ -- ubuntu-server mailing list ubuntu-server@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-server More info: https://wiki.ubuntu.com/ServerTeam