Aug 30

First question was how many of us were more than 50% virtualized, I didn’t see anyone without their hand up. I applaud us.

What are the implications of being 50% or more virtualized?
Multiple Tools
Older processes
New and more abstractions, more dynamic
More stakeholders, dis-aggregation

What capacity problem areas get amplified?
Performance triage
Reactive capacity remediations
Purchase, optimize for efficiency and cost
Collaborate & Report

Tools that we use to monitor do not always agree. I’ve experienced this issue. We all have. The other issue is the tools have alerts but those alerts are usually false positives when it comes to performance. Noisy alerts cause us to miss the important alerts. Some time ago I declared alerting “broken”. We would ignore them or we would have to spend entirely too much time validating their truth.

Staying a head of the demand in our infrastructures can be challenging. I actually accomplished this for a while with an agreement with my partner for buying blades. I had a subscription plan setup so that compute resources were shipped to on a monthly schedule.

There is an awesome slide titled “Get the right metrics”. It describes capacity and demand concepts.

Contention can show up in different ways, dropped packets, CPU ready, memory contention, latency etc. Contention is the difference between what a VM gets and what a VM wants.

Humans can feel about 10% degradation in speed.

In vCenter Operations, you should investigate “health” issues immediately. Consumed memory has no bearing on what a VM actually needs. Even as consumed goes up and down demand may stay steady.

Looking for symptoms slide is another good one. Goes through examples of under, over, and proper provisioned machines. Demand is a good thing to look at. Cannot just use one symptom to solve a problem.

A see a pattern here… Another good slide is “Mapping Symptoms to Problems”.

Get the right metrics
Change your mindset
Watch for symptoms
Putting it all together
Map symptoms to problems
Remediate the virtual way

Overall a good session of VMware vCenter Operations.

Aug 29

This session is being ran without a presentation but is being recorded and should be published later. It should be a PDF so stay on the lookout.

When building your cloud with vCloud director, make sure you create one vSphere cluster to run our management and then use the next cluster to run your cloud. VCD has 3 layers of mention, Infrastructure, Provider and Consumer. The infrastructure layer is things like the cloud networking, storage, compute, etc. The provider layer is where we pool all of this together into things like a Provider Virtual DataCenter. The consumer layer is where things like organizations go like finance, customer name, users, etc.

Inside of an organization you can create Org VDC’s. Pay attention to network connectivity for organizations. If two different VM’s are connected back to the same port group on a DVS they will be on the same subnet. vShield Edge can be used to firewall this connectivity.

vCloud hierarchical network has 3 types of networks, vApp, Org, and external are mentioned drawn.

With the API extension, you can actually provide things like iSCSI as a service via vCloud director. Now a tenant can request a service and be returned the information needed to consume it, in the case of requesting iSCSI lun, an IP address can be returned to discover LUN from.

A drawing of request to VCD to AMQP to VCO to vCenter is shown for workflow example.

There are some excellent drawings in this session being drawn live. Many are to answer questions people have on design. I have to say I like the format of this session.

I vote we make this style an official session type and potentially call the type any of the following:
Big Class
On the fly session
Experts live

They just asked the group who liked the format, I’d say 90% of attendees raised their hands.

Some cool terms for future use (definitions are my best guess lol):
Cloudify – taking something not a part of cloud and making it part of cloud. Example: let’s take this service and cloudify it.
Markitecture – using marketing terms as architecture description. Example: Put your website in “the” cloud”

Aug 28

Configuration management is nothing new. Lots of items in your infrastructure need configuration management. I think storage, operating systems, hypervisors, network, etc. vCM is a tool that can be utilized to ensure compliance with regulatory and customized compliance policies. Report on configuration drift and changes.

VMware Configuration Manager can report on and enforce compliance on many technologies. VMware vSphere can be configured in bulk across many vCenters. Yes you can use host profiles inside of a vCenter, you can actually manage host profiles with vCM across vCenters as well. Report on who made changes to vSphere, when, what changed what it should be, and depending on config enforce back to what it should be. You can use compliance policies to ensure your meeting your compliance requirements. You can make use of standardized regulatory compliance polices as well, SOX, PCI, even hardening guides.

vCM can also perform similar tasks to virtual infrastructure management with operating systems, Linux and Windows. Similar standardized regulatory compliance polices can be used and reported on. You can also apply OS patches from vCM as well as report on them.

You can use the vCenter Orchestrator plugin for vCenter Configuration Management to provide workflows to kick off tasks in vCM. If you aren’t using vCO today you should be.

VCM is part of the vCenter Operations Management suite from VMware.

I use vCM today and would like to share some additional information. vCM does use an agent installed on the machines. In the windows world this is a dcom agent and can be installed from the main console. On Linux machines the agent is installed via ssh. The licensing is per VM. The retention period for the data is 30 days by default. I would like 2 years however the database grows greatly with this (1+ TB) for 500 VM’s. You can monitor active directory user and group changes as well. When actively using the GUI it’s useful have the job monitor open and set to refresh 60 seconds for jobs and 5 seconds for tasks at the bottom.

« Previous Entries