I have recently purchased 3 Dell servers, and put myself to task to build out a new lab. My old lab was in desperate need of updating as I had long past the time when 48GB of memory per node was sufficient. The cost of memory, old or new was not even closely in line with cheap server grade CPUs that were perfect for lab servers. Today you can buy a used E7540, a low power, 12 core (HT enabled) Xeon for less than $30 (USD) from a reputable retailer. Cram two of these into an 11 gen Dell and you are in business.
So, three new (to me) Dell rackmounts, deployed as virtualization servers, and I want a simple way to view performance stats in a nice clean single pain of glass. I am not in any way shape or form looking to build fancy dashboard and setup any sort of historical monitoring. I just want to know where the performance hot spots are when my environment seems to be running slowly.
I installed Cockpit before on a laptop or two and thought it might foot the bill, especially since you could use one dashboard for multiple nodes.
So here we are going to deploy Cockpit on all three nodes, on each the steps are the same.
TestPMD is a lightweight application running in user space, utilizing ovs-dpdk, that can be used for testing DPDK in packet forwarding mode.
In this example we want to setup TestPMD on a RHEL VM running in our SR-IOV capable Red Hat OpenStack 10 overcloud. Our passthrough adapters are Intel X520s. Our plan here is to run performance tests via an external load generator.
Before we can get started we need to build a test VM.
eth0 – ssh access via admin network
eth1 – load generator private network
4096 MB Mem
150 GB Disk
Deploy RHEL VM
Your first step is to deploy your RHEL VM, configure your primary network interface (eth0 for ssh) via VM console. Eth1 needs to be up and configured to start at boot, but do not assign it an IP address. Next, register your VM with your local satellite server or with RH CDN.
Use the link below to download the “Latest Major” version of DPDK. Place the tarfile in /root on the VM and untar.
Before we can compile DPDK, we need to install a few prereqs.
When you Nova boot an instance in an OSP Overcloud deployed using SR-IOV and CPU Pinning, that instance will be NUMA aligned, meaning that its vCPUs, memory, and SR-IOV VF (Virtual Function) will all be local to the same NUMA node. Nova will not allow you to deploy a non-NUMA aligned instance in such an environment (which you might want to do when testing the cross-NUMA penalty of your hardware).
If you are looking to misalign an instance for the purpose of testing, you can use the following process.
First, ssh to the Compute node running your test instance. In the example below, we can see that the libvirt shows VM three as being pinned to odd numbered cores. On this hardware that means, NUMA node 1.
In this example, our VM has 4 cores.
# virsh vcpupin 3
VCPU: CPU Affinity
Using “virsh vcpupin“ you can move the pinned VM cores to the pCPUs local to numa node 0.
In the example below, we start my migrating vCPU 0 to pCPU 12.
If you find Toast is running very slow on your Mac and you've tried all the other fixes, check to make sure you don't have any mounts defined on your Mac as Toast will continue to try and mount them ad nauseum – thereby causing the slowness. You can check Directory Utility to see if there are mounts you forgot about.
If that's not it, it is time to embrace the built-in DTrace tools Mac has for your computing pleasure. This can help you track the calls Roxio makes when it's appears to be running dog slow.
Dtruss is very much like truss or strace – it will show you system calls made by whatever you are dtrussing. If what you are trying to figure out is shortlived, execsnoop is also your friend as it will show you system calls for any new processes as they are spawned.
Dtruss and execsnoop show when Toast starts up that automountd is always called, thus stale NFS mounts could be causing you headaches.
This my first post following my week long Redhat Monitoring and Performance Tuning Course (RH442). This course was rather intense and dealt mostly with concepts and theories rather and was not at all a how-to on performance tuning. I am going to do my best to document some of what I learned. Lets start with the documentation…
Man Pages, are the extensive documentation that comes installed by default on a linux system. They are seperated into 9 books or sections.
C Library Functions
Special FIles and Drivers
System Admin Commands and Daemons
To access a particular book you specify the book name after the man command. For example:
#man 1 printf
which is different than…
#man 3 printf
To find all the man pages related to printf, you run a the following
#man -k printf
Which gives you the output below listing all man pages related to printf. These specific ones are all on the the printf command specifically. Similarly you can use the apropos command and get the same output
To see only the man pages for printf run the command below
#man -f printf
Which will return only what you see below.
printf (1) – format and print data printf (1p) – write formatted output printf (3) – formatted output conversion printf (3p) – print formatted output