sharing files with a VM

[TheFu]

I use raw images and qcow2 for VMs and have since the beginning. If the VM is shutdown, you can mount those on the VM-host. You said you wanted a way to modify the OS for a VM from the host. I provided 2 methods. Did I misunderstand to purpose?

I'm an old MF programmer from before switching to Unix, then Linux. Actually wrote code for modified IBM 360s doing real-time control systems. IBM midrange LPARS are pretty slick on the P-series, but Sun/Oracle and HP had their versions for a long time. But none of that really matters.

AWS uses KVM now. https://www.theregister.co.uk/2017/1...oud_go_faster/

I think we're at the point where you either found what seems reasonable or not. Good luck.

[Skaperen]

Suggest you look into containers (https://linuxcontainers.org) rather than a traditional VMs. Low latency, images for different distros (https://us.images.linuxcontainers.org), all guest filestorage is file-by-file on host, different toolset options for managing via cli (ubuntu pushes LXD https://lxd.readthedocs.io/en/latest/).

i was jumping to the assumption that containers was some more sophistication/advanced virtual machine. i did that because i was looking for just such a thing and it was what i had in mind. i had heard of containers over the past few years but didn't have a reason at the time to look further. and i had heard nothing to connect them to virtualization (which they are not but there is an overlap of use cases). containers are more like BSD jail. and, as it turns out, what i am planning to do could work in containers, or perhaps a large subset could.

a container has an isolated distinct name space where one can run a suite of user land processes apart from any other. the catch is that everything runs under the same kernel. so each set of user land, which has it's own separate / directory must support that kernel. so a really old version of Ubuntu, with an older libc that doesn't understand any API changes made to the kernel, won't be able to run.

any information about which versions of user land package suites will work under which kernels would be helpful. otherwise i will just have to try it and see. i had looked into how dpkg can be run to target an initially empty directory and build a system, as an alternate way to go virtual. containers would actually make this easier since it takes out the steps to setup a boot loader and kernel(s) within the VM.

and, yes, making your own custom images is very possible, it appears. this also explains a lot of how various cloud providers are doing their pseudo-virtualization (many of them clearly mention containers as what they are doing).

this reminds me of some of the steps i had to do with the file tree back when i made a single CDR that would boott into Linux on either an Intel x86 or a Sun Sparc machine. one was made with Slackware and another was made with Linux From Scratch (from a scripted build).

i am running Ubuntu 16.04.5 LTS. it is running kernel 4.4.0. do you know what other versions of Ubuntu userland can run on kernel 4.4.0? do you think i should run a different kernel version? which version?

[Skaperen]

i have used nbd with qemu to make use of a block device space i had on another host, many years ago when i was doing some VM stuff with qemu. now days we also have iSCSI, a more sophisticated network protocol. iSCSI is also more complex compared to nbd. i'd still use nbd if the use case allowed it, for the simplicity. KISS works in configurations, too.

[KillerKelvUK]

i am running Ubuntu 16.04.5 LTS. it is running kernel 4.4.0. do you know what other versions of Ubuntu userland can run on kernel 4.4.0? do you think i should run a different kernel version? which version?

So this depends on whether you are intending to use a package manager e.g. aptitude within the container or whether you intend to build from source. As an acid test for either you can use https://packages.ubuntu.com to track available packages by Ubuntu release and then use https://people.canonical.com/~kernel...rsion-map.html to track kernel alignment to release but as aptitude will do all this automatically this would be simpler, I guess it depends how far back you need to go as to whether the repos extend that distance. Or maybe I'm not understanding your question?

[Skaperen]

I use raw images and qcow2 for VMs and have since the beginning. If the VM is shutdown, you can mount those on the VM-host. You said you wanted a way to modify the OS for a VM from the host. I provided 2 methods. Did I misunderstand to purpose?

i want to have automated file changes and i want to do that without mounting anything. what i am looking for is some means to have the files as files in the host and have the guest have a means to access them. that's why i have used NFS. now that i expect to run some VMs after not doing so for 5+ years, i am wondering if this problem has been solved, yet.

I'm an old MF programmer from before switching to Unix, then Linux. Actually wrote code for modified IBM 360s doing real-time control systems. IBM midrange LPARS are pretty slick on the P-series, but Sun/Oracle and HP had their versions for a long time. But none of that really matters.

AWS uses KVM now. https://www.theregister.co.uk/2017/1...oud_go_faster/

I think we're at the point where you either found what seems reasonable or not. Good luck.

i did old MF programming, too, but not RTCS. that's one reason i had Hercules in my VM mix a few years back. and i may do that again. but i basically just run Linux on them (still was able to play around with assembler even though it was in lower case), though i did dabble around with some old MF OSes i could get online (but have no real use for them today).

and AWS still has a lot of Xen still running. i don't know, yet, what mix of things i can run here and run on AWS.

if i can't find what i want, i'll either go back to doing NFS and/or trying some things in containers (specifically running older kernels is not the objective so containers may well end up doing most of this for me in this project)

one of the elements of what i'll be doing is running a script loop every minute to look in a specific directory for something to do. when files or directories show up, do things like compile stuff and/or run stuff and move it to another directory. the host will be moving stuff in and the guest will be doing stuff with it to see how it behaves under various versions and architectures (so i expect to run Qemu system emulations and Hercules for MF emulation). some other distros will be in the mix. i may also throw in some BSDs.