BigChief's questions

I'm trying to do a disk to disk clone of a 500GB Fedora 16 installation using Clonezilla, but it keeps using dd to copy the root ext4 LVM, and is doing so very slowly (~375MB/minute, 24hrs total copy time). I have maybe 100GB of data on this drive.

I've used Clonezilla to copy CentOS 6 box with a similar setup (1 root LVM with ext4), and it correctly copied just the files and didn't use dd to copy all sectors. Does Fedora 16 have something other than LVM2, or is the GPT messing up Clonezilla's recognition of ext4 on LVM?

OR, what should I use to more quickly clone a Fedora 16 installation?

EDIT: The LVM is definitely not encrypted.

Has anyone successfully gotten a passthrough working with the GPU portion of AMD's Fusion APUs (the E-350 is my target) on top of a Linux hypervisor? IE, I want to dedicate the GPU to one VM only, excluding all other VMs as well as the host. I know PCI passthrough can work with patches / kernel rebuilds for Xen and KVM. However, since the GPU is on the same chip, I don't know if the host OS will see it as PCI.

I know there are a number of tangential issues here, such as:

• Poor Fusion drivers in Linux at the moment
• Unsuccessful patching efforts seem common
• VT-d / IOMMU is required and (from my reading) is supported on the APU, but the motherboard may not offer it
• KVM doesn't appear to support primary graphics cards, only secondary graphics cards (described here)

However, I'd like to hear from anyone who has messed with this, even failed attempts. Fedora + KVM is my preferred virtualization platform but I'm willing to change that if it makes a difference.

EDIT: The goal is to do this for a Windows 7 guest (I know it's asking a lot). Regardless, just assume this is HVM, not PV.

Is it safe to use guestfish's zerofree option on ext4 which is inside an LVM volume? If not, what is the best approach - rebooting (inside the VM) into recovery mode and using the regular zerofree utility? I am having trouble finding any references online to using these two tools with LVMs.

I need to set the block-size of a file when I load it into HDFS, to some value lower than the cluster block size. For example, if HDFS is using 64mb blocks, I may want a large file to be copied in with 32mb blocks.

I've done this before within a Hadoop workload using the org.apache.hadoop.fs.FileSystem.create() function, but is there a way to do it from the commandline ?

I'm using pbs interactive jobs in bash, launched using qsub -I, and I'd like to run some command after launch. After this command is run, I'd like to continue to use the interactive reservation as usual.

I'm unsure how to accomplish this, because as with opening interactive ssh prompts, commands issued after the remote terminal are completed, ie qsub -I; hello.

EDIT: To clarify, in the above example (qsub -I; hello), I want to run hello inside the qsub reservation, not on the machine where I make the reservation.

I'm aware I can specify a job script for qsub, but from the man page I don't think it does what I want:

... the job is an interactive job. The script will be processed for directives, but will not be included with the job.

I do not want /home in a separate partition for a new CentOS 6 install. However, there no longer appears to be a "Review and Modify Partition Layout" in the text installer. How can I do this?

EDIT: I'm doing a netinstall, if that makes any difference.

I am transitioning from Xen to KVM, and will be using Fedora 14. I plan to use cgroups to assign priorities to VMs.

My question is as follows: If I place VMs in various cgroups, should I also put all other Domain-0 (physical host) processes in a single cgroup, which I would then assign lower priority? The analogy for Xen would be that all Dom0 processes are batch scheduled when the Dom0 "VM" is scheduled, so by placing them all in the same cgroup, I would be emulating this behavior. However, since KVM doesn't suffer from the same Dom0 IO dependencies as Xen, I could assign the Dom0 cgroup a low priority.

In other words, is there any potential performance degradation from all the background Dom0 processes, like ntpd, yum etc, competing with VMs whose resource allocations are being explicitly controlled via cgroups? What is the best practice here?

I'm transitioning from Xen 3.4 to KVM on CentOS 5.5. I need to assign CPU priorities to VMs, to ensure that one VM (vmA) always receives priority over another VM (vmB).

In Xen I could assign a high credit scheduler weight to vmA, but in KVM the only thing I can think of is nicing the process associated with vmB very high.

Since this is CentOS 5.x, it doesn't appear that cgroups are available.

Is there any better way to do this?

I have a situation where I'd like to run Hadoop spread across 2 clusters. The first cluster (ClusterA) is normal and all nodes are publicly accessible. The second cluster (ClusterB) is behind a NAT.

Nodes in ClusterA will be running both Mapred and HDFS, while nodes in ClusterB will be running Mapred without HDFS and will not be allowed to run Reduce Tasks. The master node (jobtracker, namenode, secondary namenode) will be in ClusterA.

My question is: if I start the ClusterB TaskTrackers independently without using bin/start-all.sh from the JobTracker, will this setup work? TaskTrackers in ClusterB will open their own C&C connection to the JobTracker, and should receive MapTask assignments via this connection. HDFS will be entirely in ClusterA, so all nodes should be able to access chunks fine.

The only issue I can think of is Reduce tasks running in ClusterA attempting to get intermediate data stored on ClusterB nodes. Is this a push or a pull operation? Are there any other scenarios where the NAT will cause problems?

I have a situation where I need xm console <domID> to drop into a specific user in a CentOS PV VM, say UserB. UserA was the first user created in the VM, and whenever I xm console <domID> I get UserA. I tried deleting UserA but then get the following when I connect:

User not known to the underlying authentication module

How can I change the user which xm console <domID> connects to?

Looking online, most of the serial console tweaks for CentOS are in grub. I'm using the kernel line in the config and therefore am not using grub.

We have ~100 physical nodes which can host a large number of VMs apiece. We need to set up dhcpd on one machine to distribute addresses in a predictable fashion based on MAC address.

We'd like to go with a scheme such as:

Physical host address: A.B.255.X
VM address range: A.B.X.Y
VM MAC range: 54:52:00:00:x:y

where

x = Hex(X)
y = Hex(Y)

I found this post which would let us set up A.B.X.* pools based on MAC prefix 54:52:00:00:x:*, but this has two drawbacks:

1. Each pool is dynamic, so no direct y<->Y mapping
2. We'd have to explicitly name all the pools in the dhcpd.conf (~100 of them)

Is there some way to succinctly specify a scheme like this in dhcpd.conf?