Vcp4

Mishra

I'm going for the VCP4 certification with a co-worker. Oh boy!

Next week, 12/1/09, I'll be taking the training class. This week I hope to really dive into Vmware4 before I take the actual class... I always like classes to be more of a refresher for every subject's top level while learning in depth skills... Like the MCSE, I want to know what DNS does, have some basic experience with it, but learn what itertive queries are in class.

Anybody have any links or material that would be useful before taking the class?

Thanks!

MentholMoose

Register for the vSphere Evaluation and go through the VMware vSphere 4 Evaluator's Guide. I haven't done this myself, but I did do the VI3 Evaluation and I learned most of the basics, which sounds like what you want.

Also, I'm looking into take the class, so post your impressions and opinion of the class if you don't mind.

Mishra

Thats cool, I'm setting up a lab now.

Any other suggestions?

MentholMoose

Mishra wrote: »

Thats cool, I'm setting up a lab now.

Any other suggestions?

Not really, just enjoy the class!

meadIT

The class (at least mine did) just steps you through the high level stuff. It doesn't really go into the amount of detail you need to know for the exam. You'll need to do a lot of hands on and reading/memorization for the actual exam. You don't really need to know anything before walking into the classroom.

kalebksp

meadIT wrote: »

The class (at least mine did) just steps you through the high level stuff. It doesn't really go into the amount of detail you need to know for the exam. You'll need to do a lot of hands on and reading/memorization for the actual exam. You don't really need to know anything before walking into the classroom.

I agree, the class isn't intended to be exam prep, it's essentially VMware for beginners. If I had the choice I would have preferred to do the FastTrack.

Mishra

Thanks for the info.

I'll just hit up as much as I can. I've done some studying... I'm trying to mix in the CLI commands with every GUi change I do.

Mishra

First day done... Very smart instructor. He said he was like the 13th registered vmware trainer ever... And travelled a ton doing things directly for vmware.


Class was pretty slow... The instructor teaches very well for all around students but learning how to install esx in 2 hours and vcenter server in 2 hours is insane... I think the last 3 days will be great.

dynamik

Yea, the first day is mostly marketing

I was miffed about the course requirement since I prefer self-study, but I thoroughly enjoyed the course overall.

astorrs

Mishra wrote: »

First day done... Very smart instructor. He said he was like the 13th registered vmware trainer ever... And travelled a ton doing things directly for vmware.

Who's your instructor?

Mishra

John Davis... He works for new age technologies which is a local company to Louisville. He worked directly with vmware through new age his entire time with VMware which is about 6 years of experience.

Mishra

I wrote some completely random notes down... It could have been things I didn't know, I wanted to remember, or I just felt like writing down... Also a couple of items may have been incorrect because I didn't hear him clearly.

VMware training: Day 2

Distributed Switches
· If the ESX hosts are added to a vcenter server, then build large dvswitches if you have many hosts you have to manage
· Attach the ESX servers to vcenter server to the dvswitches
· You can attach a policy to the dvswitch instead of attaching the policy to every esx host in your environment
· The management of dvswitch is controlled by vcenter server, however if the vcenter server goes down then you can’t manage the dvswitch but data is still being transferred
· The dvswitch is a switch available across the entire datacenter to use
· The dvswitch is basically an ‘invisible’ management source, it still builds the vswitches on the esx hosts like normal, or assigns the policies to the esx hosts like normal… so that means that the vswitches still only talk across the same esx host, it doesn’t travel up to the vcenter server and back down because the dvswitch is an ‘invisible’ management source
· The datacenter is a boundary for dvswitches, you cannot connect hosts from seperate datacenters to one dvswitch
· You can only have up to 32 uplinks in a dvswitch
· Virtual adapters is kmkernel ports and service console ports associated to it in a dvswitch
· You still have to create vmkernel ports on all the esx hosts, but if you want to make a global change, you can change it on the port group in the dvswitch
· Cisco Nexus 1000 dvswitch is a third party dvswitch that uses the APIs… this might support QoS or dynamic vlans which VMware doesn’t support
· Private VLANs are only available on dvswitches
· Private VLANs are a way to route in VMware without having a layer 3 device
Policies
· If you want to put a sniffer on your network, you have to enable promiscuous mode
· You can traffic shape per portgroups


Storage
· You can only use NFS in a NAS environment
· VMFS has vma.vms | vma.vmdk which has metadata/text/small | vma-flat.vmfs which is data/pre-allocate
· Thin provisioning doesn’t pre-allocate the vmdk files, it grows as the virtual machine grows
· Raw device Mapping is a difference in your flat file. You can present an entire LUN to a virtual machine and format that LUN with your vmdk-flat file… Your vmdk and vmx files are still stored on your regular VMFS datastore… You may want to do this because of clustering or easy of use.
· To use Vmotion DRS or HA, you need shared storage
· Vmhba:C:T:L is Controller/Target/LUN
· iSCSI supports 10g networks
· IQN name is a part of the iSCSI initiator name
· There is a purpose to have different iSCSI addressing instead of using regular IP/MAC addressing.
· Different iSCSI verdors might give you separate targets depending on how that vendor presents the LUNs
· You can grow datastores instead of spanning in version 4. You cannot shrink a VMFS datastore. You have a 2 TB mins 512B limit in your VMFS datastore
· You can extend 2 TB volumes up to 32 times

Virtual Machines
· .nvram is used to the machine BIOS .vswp is swap and .vmsd is the VM’s info for snapshots
· You can only go up to 255GB of RAM for a virtual machine (check at vmware.com)
· Each virtual NIC gets a MAC address
· Thin Provisioning is a new feature
· 2003 defaults to LSILogic and 2000 defaults to the BusLogic virtual adapter
· Windows normally defaults to e1000 but you can change up to better performing vmxnet or vmxnet3
· You can change your virtual machine version in order to import a VI3 VM (version 7 is the latest)
· OVF is open VM format
· You can convert Virtual Machines to templates through ‘Convert to Template’ option
· Host and Cluster view does not show templates, use VMs and Templates
· VMware gives you the option to run ‘Guest Customization wizard’ to ‘sysprep’ the virtual machine to make it unique
· To run guest customization wizard, then one time you need to download the sysprep files from Windows. If the option to customize is greyed out, it means you haven’t downloaded the sysprep files yet.
· You can deploy templates across datacenters, it is NOT a boundary
Standard Vswitch
· You need a vmkernel port to connect to an iSCSI network


Notes
· Virtualizationinformation.com/docs/Supplemental_documents.htm
· 3 purposes of vmkernels are: iSCSI, vmotion, and NAS/SAN/NFS connections



Challenge
· Script a change to a vswitch across multiple hosts.. This is to prove that you don’t need dvswitches in certain environments

Mishra

Do you all think I should have the entire 'configuration maximums' document memorized for the exam?

kalebksp

Mishra wrote: »

Do you all think I should have the entire 'configuration maximums' document memorized for the exam?

No, it probably wouldn't hurt to memorize the more important ones (RAM, CPUs, etc), but I wouldn't worry too much about things like the max file size for specific block sizes of VMFS2. Based on what I've read the maximums were stressed more in previous VCP exams than VCP4, I don't know for sure though since I started with VCP4.

astorrs

I agree, just focus on the big ones, CPU, memory, etc plus cluster size, nfs volume/LUN count, etc

mzgavc

I also had John as an instructor.

Great guy.

meadIT

kalebksp wrote: »

No, it probably wouldn't hurt to memorize the more important ones (RAM, CPUs, etc), but I wouldn't worry too much about things like the max file size for specific block sizes of VMFS2. Based on what I've read the maximums were stressed more in previous VCP exams than VCP4, I don't know for sure though since I started with VCP4.

Also agreed. But you should know the max file size for specific block sizes for VMFS3. The maximums don't seem to be stressed much, but they are pretty much gimmes on the exam if you know the major ones.

Mishra

Here are my other notes

VMware training: Day 3

OVF
· OVF is an open VM format and it is a way to package VMs in order to create a VM appliance
· You can directly import a VM appliance to different VMware products like Server/Workstation/ESX
· You can VMs you built and package them in OVF format to deploy in other environments
vAPP
· It might be a case where you need to stop/start/package/backup an application so you may need to start/stop VMs in a certain order because of a product
· You can take a set of VMs and package them as 1 package to maybe redploy that package in another environment
vSheild
· Didn’t come directly with enterprise
· This is a VM appliance that is easy to deploy and you can configure it through the client with a plugin
· This is a firewall
vmSafe
· Different technologies and services and vsphere will recognize
· Can help you lock down and protect your environments
· This will allow you to do things like intrusion detection within the vswitch or other things
vCenter Converter
· You can enable the plugin for converter on certain vsphere clients
· You can import other virtual machines like from vmware server/workstation
· It could be a ghost or acrois that you import as well
· It might be the faster way to move a VM from esx to esx server, you can move it and shrink the disk during the move using converter
· Approximately 300MB of disk space is needed for all components for Converter
· Hot cloning needs 350MB of memory is required on source physical machine
· Cold cloning is at least 264MB of memory is required on source physical machine; 364MB or more is preferred
· An agent has to be installed on the running physical machine to convert correctly
· Converter transfers data from the source directly to the ESX host
· Choose ‘import machine’ to start vm converter
· Watch for applications that depend on MAC addresses/video cards/serial numbers/hardware characteristics
· You have to disable Windows firewall in order to hot-clone a system
Guided Consolidation
· This tool will help you understand if your physical machines can fit on the ESX servers you are planning to implement in your environment
· It adds a data collector to the existing physical server
· This is located under /home/solutions and applications/guided consolidation/vCenterServer
· It will use perfmon in order to gather statistics
· There is a better product called VMware capacity planner/CapacityIQ… guided consolidation is mostly used for small business
Modifying/Managing Virtual Machines
· You can increase the size of a virtual disk while it is hot
· You can suspend either through vmware or let the guest operating system to handle the suspension
· You can check a box to make it go immediately into BIOS for ease of use
· Paravirtualization lets the guest operating system be aware that it is running inside a virtual machine rather than on physical hardware
· You can store the swap file by defaults, with the VM, or on the ESX host
· When you make a snapshot, it creates static files, however any changes that are created are put on a delta vmdk file, it also makes a new metadata.vmdk file
· You can snapshot RAM as well, however it may be disruptive to end users
· Make sure you remove other snapshots after they are made once you think your changes you made that caused you to created the snapshot in the first place
Storage vMotion
· Change host is regular vMotion, if you use change datastore, you are using storage vMotion
· Perform vMotion during off-peak hours
· Ensure that source host has access to both source and target datastores
·
Access Control
· Default permissions on the vcenter server is the group administrators has full control

Virtual CPU and memory concepts
· vmmemctl is called ballooning
· Each virtual CPU will map to a HEC or hardware execution context
· Each virtual CPU per VM will map to a logical CPU, if you have hyperthreading on, then each VCPU can point to 1 HEC on each core… Meaning 1 physical CPU can have 2 HECs and a dualCPU VM can sometimes point to 2 HEcs on 1 physical CPU but VMware normally does not point like that unless necessary. It understand hyperthreading
· Service console only stays on the CPU 0, the VMs will balance load between all CPUs
· Cpu ready time is an indicator of CPU contingence… or a CPU will wait for another CPU to continue processing
· 5%+ of CPU ready time means you could start seeing problems
· Every 20ms, it could change the VCPUs to map to another HEC
· The partner VCPU’s will also be mapped when the first VCPU is mapped.. this happens at the same time
· You can over commit RAM on machines
· The VMkernel will reclaim memory because it only maps the same page once during in physical RAM. You can possibly say 30% in RAM with this ability.
· Swap file for RAM is last resort for the VMs to save it’s RAM in
· By default the VM won’t balloon away 65% of the memory. You can reserve(guarantee) RAM per VM.
Monitoring Resource Usage
· The disk queueing graph is not in performance console, you have to use the CLI. Esxtop is used to find this
· Network performance, you look for packets dropped to see ifyou have a problem
· vappSpeed is an addon for vcenter server that gives you more performance options
· visioncore/foglight is other software
Scalability
· Per VM, go to the resources tab, and change resource settings.
· You can change Limit/Reservations/shares in order to change settings
· The CPU limit is unlimited by default
· Shares are priority under contention
· By default, the VMs have 1000 cpu shares. If you made 1 VM have 2000 shares and 1 VM had 1000 shares, then the 2000 shared VM would be twice as likely to be chosen
· As the above scenario, you have 1 GHZ, VM1000 would get 350MHZ and VM2000 would get 650MHZ
· Processor affinity will force a VM to use a certain CPU
· You can make resource pools, like prod/test/dev
·

Mishra

VMware training: Day 4

Storage and Network Management
· Mutlipathing is built into ESX
· You can change which multipath is active, but there isn’t many things to change in multipathing
· Round Robin – a multipathing policy that performs load balancing across paths
· Path selection is normally dictated by what storage you have.
· Availability – MRU and Fixed
· You should have a total of 4 combinations in your multipathing screen, 2 for SPs and 2 for HBAs
· PSA – Pluggable Storage Architecture, this is a third-party way of developing for PSP (path selection policies)
· NMP – Native Multipathing Policy
· If you have 3 physical adapters, and you connect a virtual adapter… It will be mapped to one of the 3 physical NICS even when teamed
· A virtual NIC chooses it’s physical NIC by taking it’s portID and dividing by 3 (or how ever many physical NICs you have) then assigns by the remainder
· One option is you can assign MAC based load balancing, which means it’s uses it’s MAC address and divides by physical NICS
· VMware does not do smart load balancing, if a node is saturated, it still may be used when a VM is powered on
· 3rd option is IP hash, it takes the IO from 1 virtual adapter, and it distributes across all physical NICs… it does this by taking the IP address from each client and divides by the physical NICs. CON, the physical switch needs to do etherchannel to perform IP hash. It may apply that all 3 NICS have to attach to the same physical switch.
· You can perform an active/standby port group in order to ‘manually’ assign a VM to a physical NIC
VMotion
· VMs have to connect to the same storage in order to perform VMotion
· Vmotion does an rarp in order to change the MAC address in the network after a vmotion
· VMs must meet the following requirements: It cannot be connected to an internal vSwitch (vswitch with 0 uplink adapters) ||| must not have a connection to a virtual device (like a cdrom) ||| cannot have CPU affinity configured ||| must not be in an MSCS cluster ||| swap file much be accessible by destination host ||| if it uses RDM, it must be accessible by destination host ||| visibility to all storage ||| gigabit Ethernet backplane ||| access to the same physical networks ||| compatible CPUS: CPU feature sets of both the source and destination host must be compatible, some features can be hidden using enhanced vmotion compatibility (EVC) or compatibility masks
· You can hide NX/XD flags from guest for vmotion
· Use VMware bootable CPUID utility to find out if the CPUs are compatible
· You can tell quickly if you can Vmotion a VM by looking at the maps tab and see if it doesn’t have an X
· DPM is Distributed Power Management
· DRS is distributed Resource Scheduler
· DPM requires DRS
· EVC is Enhanced VMotion Compatibility
· EVC will look at a cluster and will mask away certain dependencies so you can use vmotion/drs… this can cause performance reasons because it is not using a certain feature of the CPU
· You can choose to put your swap files on the local drives of the ESX servers using VM swapfile location
· You should make cluster based resource rules instead of host based for best practice/ease of use
· You can create rules in DRS to avoid 2 ‘clustered’ servers from being on one machine. You can use ‘seperated virtual machines’ or other settings
· To service a host in a cluster, you must place it in maintenance mode
· If you are using DRS and you place it in maintenance mode and you are fully automated, it can move the VMs and balance automatically for you
High availability and fault tolerance
· VMware fault tolerance is new in vsphere 4
· If MS clustering is a good fit in an environment they may keep MSCS.. Else you could use VMware fault tolerance
· High availability and fault tolerance can work with almost everything
· You set different properties in VMs for MSCS
· The data that you share for MSCS should be a raw device mapping, you should add a second virtual scsi adapter and change bus sharing
· You can MSCS a vCenter server or use VMware vCenter heartbeat… this is a new feature in ESX4
· VCB -> VMware Consolidated Backup
· VCB framework would allow backup software to work with your VMFS partitions
· VDR -> VMware Data Recovery
· VDR is based on the VMware vStorage APIs for Data Protection… It is an agentless, disk-based backup-and-recovery solution for virtual machines, based on a virtual appliance
· VMware HA is a quick reboot of VMs after an ESX host failure
· HA runs without vCenter Server being up… this uses a heartbeat between ESX hosts on the service console
· You want to avoid host isolation which means that service console is not running for some reason which means that HA isn’t getting a heartbeat…. You have certain options on what is should do… You can possibly shutdown the guest operating system, wait, then let it run on other ESX servers… You can also shutdown the operating systems hard and then turn it on the other ESX servers
· You should probably add more service console ports in order to avoid ‘no availability’ which could mean that for some reason, all of your service consoles got disconnected (like vlan is screwed up)… You want to be on different networks probably
· Shared storage, Make sure DNs resolution is on, make sure time is synced for HA to work… This is like the requires for VMotion
· You can set a tolerance level for host failures
· Admission Control tells you how many VMs it can hold for HA
· You can reserve a percentage of total cluster capacity for HA
· ESX4 gives you a new option to enable VM Monitoring in VMware HA
· VDR -> VMware data recovery is a VMappliance
· You can perform disk to disk backups with VDR
· Install a plugin in vcenter client to use VDR
· VCB doesn’t give you a direct way to restore file

Update Manager
· Update manager is installed through the vCenter Server installer
· Update Manager creates it’s own account, it only uses the local admin account once
· Normally you setup a separate database for Update Manager
· You can use this tool to patch your guest operating systems
· Remediate is to install patches
· You have to scan to see if your patch baseline has updates it needs on all the hosts your configured your baseline on
· If you have DRS it will put the host into maintenance mode and all the VMs will move off to the other host. DRS has to be fully automated
·



Notes
· You can’t have dvswitches configured and change vcenter server hosts

meadIT

When it comes time for you to take the exam, vReference has a couple of good resources, one is a condensed version of all the VMware PDFs with important things bulleted.