nponeccop's questions

Short version:

I want to copy a thin provisioned VMFS .vmdk from the free ESXI to a NAS for backup. How I do it? Are there any way which works out of the box? Naive copying doesn't work because it copies full size instead of actually provisioned size.

Note that the particular .vmdk format relies on VMFS sparse files. However there are other thin allocated image formats, including .vmdk used by Workstation, and .ovf. So exporting the images to those formats would work just fine.

Long versions:

I have a lab ESXi 6.0 server (I mean for experiments, not production) with a couple of VMs there. And I want to copy .vmdk files from the ESXI to my QNAP NAS for backup purposes.

However I don't want a full backup solution. I just want to keep the QNAP offline and backup like twice a year. At this pace the backup solutions, especially free solutions, don't survive well - they tend to rot due to dropping OS support, changing formats, forgetting passwords and so on.

So I want something with close to zero setup and maintenance. Recovery performance isn't important.

I have an up to 400 GB thin volume with 40 GBs actually used.

However a simple .vmdk download from the web-based Datastore Browser downloads full 400 GBs which doesn't look right to me.

The same goes with scp/sftp.

The only solution I have found so far is to lzop the image before downloading.

So the question is, are there any better ways which aren't overly complicated? For example Veeam requires remembering account password, using "corporate" email and downloading some 4GBs of their backup solution.

What mechanism Hyper-V use to signal a Linux guest to gracefully shut down?

Somehow CoreOS does gracefully shut down, but another Linux doesn't (it is just abruptly powered off)

I want to connect a Linux VM running in VMWare on a Windows 7 laptop to an enterprise network using host-only configuration.

I was able to setup bidirectional routing between "host-only" VMWare network and the enterprise network. If I add DNS servers to /etc/resolv.conf and a default route then there is full bidirectional connectivity. Traffic flows between enterprise network and host-only network, and NAT is performed later by the enterprise WAN gateway.

However, out of the box no default route is passed over DHCP, and the DNS server provided doesn't work.

I have the following topology:

VM - 192.168.4.3 -- host only network -- 192.168.4.1 -- laptop VMWare host -- 192.168.1.4 -- enterprise network -- 192.168.1.1 -- WAN Gateway -- Internet

DHCP sets up 192.168.4.1 (the laptop interface facing host only network) as the DNS server on the VM host. VMWare has a DNS proxy which works fine in NAT mode but it refuses to work in host-only mode.

So I have 2 questions:

1. How do I configure VMWare so it sets correct DNS (either DNS known on host, e.g. 192.168.1.1, a working VMWare DNS Proxy or a good static DNS e.g. 8.8.8.8/8.8.4.4 Google Public DNS)

2. How do I configure VMWare so it sets a proper default route to 192.168.4.1?

I didn't found the proper forum - so move the question if you find a better place for it.

We have 10 houses scattered in the area 500 meters in diameter, and have a tiny DIY CATV network providing houses with 2 analog TV channels (don't laugh it's third world).

We use cheap CATV signal boosters/line amplifiers at some intermediate houses to get the TV signal to the farthest houses.

Now we want to setup an IP network between the houses, so what is the lowest cost solution for that? We won't need more than 1 megabit, so basically speed doesn't matter.

The lowest cost solution seems to be "ethernet switches hanging on trees", but it's not reliable.

DOCSIS seems too powerful for our tiny setup, so we consider DSL (VDSL/SHDSL, also it seems there are 8-port ADSL DSLAMs for $600), EtherLoops ($160 per modem pair) and newer EoCNA standard. The EoCNA seems attractive because of reuse of our CATV cables.

The question is, if we decide to deploy ITU G.9954 Ethernet Over Coaxial Network (EoCNA), what are the requirements for our signal boosters? Do they need a special capabitity to pass the backward signal? Are there any good white papers/docs about the ITU G.9954 technology?

What else technologies can we consider, given we targeting $300 per house with up to 10 houses?

Some houses may be rather close, so we think of mixing long-range and short-range techs (e.g. wifi repeaters or plain ethernet between some houses) if that can lower the cost.

Arch Linux doesn't have sendmail installed by default. So I get the following messages in my syslog:

Sep 15 13:16:01 zorro crond[18497]: mailing cron output for user collectors sh cronjob.sh
Sep 15 13:16:01 zorro crond[18497]: unable to exec /usr/sbin/sendmail: cron output for user collectors sh cronjob.sh to /dev/null

What is the recommended way to fix this default behaviour so actual messages are sent?

heirloom-mailx is installed and capable of sending email messages using SMTP. Is it possible for crond to use mailx to send notifications?

Is there any drop-in replacement for sendmail that sends using mailx? Sendmail is not even in the repositories.

I'm trying to install Arch Linux under Xen as domU (Guest) from Arch Linux official 2011.08 x64 Core iso image. My web hosting offers a SolusVM control panel with ability to install custom ISOs, so I have full control over guest, but not over host.

I can install from 2011.08 i686 Core iso image - it's x64 flavor that fails.

The installer fails to find both network adapter and disk drives.

The hosting support says they have x64 guests working fine, but Ubuntu 11 minimal x64 fails to detect network adapter there (under the same Xen dom0) too.

When I compare lsmod | awk '{print $1} | sort from 32-bit and from 64-bit setup shell, I see that 32-bit setup loads i2c_piix4 instead of i2c_piix and also loads 8139cp, 8139too and ata_generic modules. But xen_platform_pci kernel module is only loaded by the failing 64-bit setup.

I cannot finish installation because it is impossible to create partitions without available disk devices.

If I pass xen_emul_unplug=unnecessary option to the kernel from syslinux, installation works fine. It seems that the kernel sees that paravirtualized devices are available, unplugs the emulated devices but fails to plug the paravirtualized devices.

• Is this a bug, a misconfiguration, or by design?
• If it's not by design, what causes it and how to implement a proper fix?

I bought a VPS but the hosting company doesn't have a prebuilt image for Arch Linux x64, so I have to install from ISO using VNC in SolusVM.

The installation works fine, the box (Xen Guest) is bootable, but after an upgrade (pacman -Suy) kernel26 gets replaced with linux-3.0 and stops booting.

The message is:

Root device /dev/disk/by-uuid/{uuid} doesn't exist. Attempting to create it.
ERROR: unable to determine major/minor number of root device '/dev/disk/by-uuid/{uuid}

Only by-id and by-path folders exist in /dev/disk if checked from emergency shell, and the only device is QEMU CD-ROM.

How do I fix it so linux-3.0 boots under Xen?