VMware has evaluation license for ESXi servers. After 60 days, it expires and you have to apply a paid or free license to continue. Technically, there is a trick to reset the evaluation key by deleting two files (/etc/vmware/vmware.lic and /etc/vmware/license.cfg) and rebooting the server. It’s of course not complying with VMwrae license terms. Under some circumstances like training lab, it may be OK. Make sure to consult VMware on this if you are not working for VMware. But wait – if you are working for VMware, do you need evaluation license? In his reply to my tweet, Duncan mentioned he never saw license expiration.
Although VMware ESXi supports common Linux commands, its implementation is based on busybox. Some of the commands are not supported, or functionalities are reduced. Setting time is one of the cases.
To find out the current time, you issue the following command:
~ # date Sun Jul 13 10:50:59 UTC 2014
Now if you want to use the same date command to change the time, it’s a different story. In fact, the help of the date command works as normal, but when you really type in full command, you’ll see it won’t work.
In my last article, I discussed on the format requirement for Java APIs and how I found out the root cause and its solution. Even more mysterious is the format requirement of VMware vCenter as I discovered in another VMware related project, in which I needed to register an extension with vCenter and set up its certificate.
While debugging a vSphere Web Client plugin project, I found it’s not easy to refresh the services with the Virgo server which acts as the back end for the plugin GUI but as client for the vCenter server. Packaged as OSGi bundle, it’s supposed to be easy to reload the service. Mixed together with various components in the plugins, however, it’s sometimes not quite straight forward for the re-deployment for updated code. Here is a brute force approach I found while playing with it.
Since I left VCE four months ago, I have been working intensively on a commercial version of the open source vijava API supporting all versions of vSphere APIs (5.5 is the latest). If you have used the open source API, you know the vijava is much faster than other alternatives. Since its debut, it has been used in many commercial products from companies like Cisco, EMC, HP, etc.
As reported in the community, there were quite excitement about the open source of the pyVmomi, the Python equivalent of vijava API. It was heatedly debated whether to open source the API even when I was working at VMware years ago. One camp of people thought it should be open sourced and even supported as Web Service SDKs; while the other group didn’t think it’s mature and would cause a lot of trouble in so doing. So it didn’t go anywhere in the past few years.
I just got into a very interesting problem recently – the vim-cmd does not work as expected when used for renaming a datastore in vSphere.
What is the problem exactly?
The following command, for example, should change the name of a datastore from datastore1 (which is the default datastore name) to doublecloudDS.
# vim-cmd hostsvc/datastore/rename datastore1 doubecloudDS
After the command is executed, there is no error message reported. But the datastore name remains the same as shown in either the vSphere Client or using the following command:
My article “Run esxcli Command in a Web Browser: Another ESXi Hack” got quite some interests from the community. Although it works, I am not quite satisfied with the fact that the real esxcfg-info.cgi is disabled to run the esxcli.cgi.
With the vSphere Web Client, VMware has really made the system complicated and slower. The extension mechanism is more flexible, but forces developers to use more libraries/frameworks/languages, therefore represents a much deeper learning curve than before with the Web based plugins. Installing and configuring the development environment itself could be intimidating for some developers. That is why I wanted to avoid it as long as possible, until I got a consulting project that may involve developing plugin for the Web Client.
In my recent consulting projects, I really got into a lot of scripting either command lines or Python with ESXi management. As I mentioned the hidden HTML formatter in esxcli command, you may have speculated what could the usage. The answer is simple: Web. But it’s not quite clear how it can be used. That’s where my curiosity started.
For those who run ESXi on a virtual machine, it’s a great news that VMware has released VMware Tools for nested ESXi as a fling in VMware Labs. Why? With the VMware Tools, you can get guest OS (really the ESXi here) information, like the IP address directly. It may sounds trivial as you can see the IP address from the virtual machine console of a virtualized ESXi. But for automation, it’s pretty hacky to get it programatically. Some people may wonder, “why not run commands via SSH?” It’s true that it’s easy to get the IP by running esxcli command, but you have to get IP first before running the command. With the VMware Tools, you can easily get the IP from vSphere Java API as would with any other normal virtual machines. Even more, you can also run commands like vim-cmd/esxcli in the virtual ESXi via APIs.
Besides the vim-cmd command I covered earlier, there is another powerful set of commands in ESXi – esxcli. As you can find from the help of the command, it covers 10 namespaces and drills down several layers down. The typical operations with the namespaces are get, set, and list. If you are familiar with REST API, you can think of the bottom level namespaces are resources.
If you have read my previous article on the vim-cmd, you may have realized how handy it is, especially when it comes to manage virtual machines. There is however a pretty challenging problem to use it – for most commands for a virtual machine, it requires vmid which is an integer that uniquely identifies the virtual machine in the context of an ESXi server. It’s like primary key in SQL database to locate a record (virtual machine instance) in a table (virtual machine type). For people who are familiar with vSphere APIs, the vmid is the same as the value of ManagedObjectReference value of a virtual machine in ESXi. Because most administrators who use commands are not necessarily familiar with vSphere API, it doesn’t help much.
vSphere Client and vSphere Web Client allow administrators to download system logs from different ESXi hosts with choices of predefined groups of information like System, Storage, Network, UserWorld, etc. Under each group, there could be multiple types. For example, under the UserWorld, there are HostAgent and ProcessInformation.
As a powerful virtualization server, ESXi has a built-in SSH server even though it’s not enabled by default. That is what most system adminstrators use to remotely run commands there. ESXi also has a built-in SSH client so that you can ssh to other servers from ESXi. To use SSH as either server or client, you need to open up firewall. You can use vSphere Client to do it ( on host’s Configuration tab, check out the Security Profile in Software section), or simple with command line as follows.
It’s not a secret that VMware has a private Python API or so called Python binding for vSphere API. If you haven’t heard about it before, no worry. Here is a link to Hostd General Architecture. Somehow it’s not publicly released as a product for customers or partners. Over the years, I only heard a big bank uses it for internal IT automation. But it’s super easy to get it if you want – it’s part of every ESXi installation. Just check it out at /lib/python26-visor.zip if you SSH to your ESXi box. Update: in ESXi 5.5, look at the /lib/python2.6/site-packages.
As I introduced in the article on vim-cmd commands, you can use a very simple command as follows to create a new virtual machine. Alternatively, you can ignore the path after the datastore and provide only datastore name (The [ and ] are still needed).
# vim-cmd vmsvc/createdummyvm testVM “[datastore1] testVM/testVM.vmx"
Other than the name and configuration file path in data store, there is no additional information provided such as the size of the disk, memory capacity, etc. Normally, you have to go through a wizard of several pages to create a new virtual machine.
While preparing for my home lab, I have created several virtual machine templates. Here are a few tips I found useful to smoothen the process and make your virtual machine templates easy to be deployed than otherwise.
Install VMware Tools
As you may have known, VMware Tools brings many features to the table, for example,
Significantly faster graphics performance and Windows Aero on operating systems that support Aero Copying and pasting text, graphics, and files between the virtual machine and the host or client desktop Improved mouse performance Synchronization of the clock in the virtual machine with the clock on the host or client desktop Scripting that helps automate guest operating system operations
Wait, it does not even mention APIs. For Guest APIs in vSphere 5.0 and later to work, you must have VMware Tools installed in your virtual machines.
To understand the ESX Agent Manager API, we have to first explain the Agent, which is essentially Agent Virtual Machine. The agent virtual machine can be hardware drivers for your ESXi server, or simply software, i.e, virus scan, that should be deployed on each ESXi. They could have been designed and installed directly on ESXi via VIB, but it would increase the risk of destablizing ESXi due to access to lower level APIs of ESXi. To lower the risk, the driver VM idea came up – if the driver VM crashes the ESXi is still solid even though some service may be affected.
Command lines are very important for system administrors when it comes to automation. Although GUIs are more likely (not always as I’ve seen too many bad ones) to be more intuitive and easier to get started with, sooner or later administrators will use command lines more for better productivity. Check out DoubleCloud ICE if you want the best of both GUI and command lines.