CMD – Massdelete files

Not sure what happened, but thousands of small files piled up in the path below.

My google-fu wasn’t sufficient enough to find out what files in that folder actually did. An article from Kofax said that just deleting files could be troublesome. But also found another article that stated that he deleted all files older than 30 days and hasn’t had problems yet. So I dare to do the same. The command was found on a blog, so all credit goes to this guy.

forfiles is a very nice command that iterates through the files in a folder according to its parameters. /D -30 iterates through all files more than 10 days old. attrib -s takes off the System attribute, which is needed for DEL to work. The echo is there so you can see that it is doing its job.

Extend disk with LVM

Here is a quick walkthrough showing you how to expand an LVM volume or partition in Linux by first resizing logical volume followed by resizing the file system to take advantage of the additional space.

Note: In this example, we are working in Ubuntu, some commands may differ in different Linux distributions.

DISCLAIMER: Make sure that you have proper backups in place before starting out with the resize procedure on your VMs! Create any backups necessary to ensure that if something goes wrong you can always go back to a previous working state. Losing any information or wiping out your disks is all your fault if this happens. Using this procedure is all at your own responsibility.

If you are unsure about LVM and its components I will suggest you read a bit upon it. There are tons of articles, for example on Digital Ocean.

Process

This process can be easy to do with LVM as it can be done on the fly with no downtime needed, you can perform it on a mounted volume without interruption. In order to increase the size of a logical volume, the volume group that it is in must-have free space available. It goes as follows.

  • Add more space from the hardware level. Either raid controller, or hypervisor.
  • Resize your partition to contain the extra space
  • Resize the PV in LVM
  • Expand the LV in LVM
  • Resize filesystem

To view the free space of your volume group, run pgdisplay command as shown below and look at the “Free PE / Size” field, in this case non-free.

Since we have added more space from the hardware level need to grow the partition. This is done by deleting it and create a new one that is starting from the same sectors but where the new partition uses more sectors than before. In the output beneath we can see how to extend the partition.

The disk partition is now extended. We need to inform LVM to grow its PV

Now we can verify that there is available space to grow LV.

Lastly, we grow the LV and extend the filesystem.

Conclusion

We have now successfully expanded a file system and corresponding LVM logical volume without any downtime. This was done by first expanding the partition of the disk, then the logical volume and finally performing an online resize of the file system.

No servers were harmed doing this procedure 🙂

Cisco ASA HA member swap

The old Intel C2000 bug is still among us. Not all devices have been affected, yet. For me, this is the first. So here is my 2 cent on how to remediate the cluster with an RMA unit.

https://www.cisco.com/c/en/us/support/docs/field-notices/642/fn64228.html

Process:

  1. Label all cables from the faulty unit, this is for you to not worry that any cables are mislocated when you are swapping the unit. I also disabled the links for the switch so that I with a calm can plug them in again without fearing that something will go wrong. Better safe than sorry.
  2. Make sure the firmware level on the new device is the same as the existing member. Firmware up/downgrade is needed.
  3. Backup the config of the HA member still running. Just in case something bad happens.
  4. Give an IP on the HA link interface, from this point it should find the existing member and start replication the config.

Firmware upgrade:

I have booted the unit and linked it only with a console cable. I can now see the firmware version and it needs an upgrade.

I have prepared a USB disk format in FAT and downloaded the matching firmware from cisco.com. When you plug in the USB key to the ASA you should now see a disk1 where you can copy from. If you want to see disk1 content issue command “show disk1”.

We now copy over the files and make the system boot the new firmware.

After reload, the system is now up and I can confirm that it has booted on the new firmware.

Joining the HA cluster

We now verified that the two ASA firewalls are on the correct firmware level. Now connect all the cables to the firewall, on the switch side all data links are administratively down, the HA link between the two ASA is a dedicated link. And those are the links we are now going to configure.

You can grab the lines from the existing member that are actively running. If you don’t have the failover key, you can also reset this on on the primary/existing member.

The new ASA is now ready to contact the primary member of the cluster and start the replication. In my case, the interfaces for HA were administratively down. So we are now going to enable the link and enable failover.

If something in the config is not ok, missing files or other is listed and you have to remediate this before you again can try to enable HA with the “failover” command. In the output beneath you can see what happens when the failover command is enabled with success.

You can now check failover status and see if the standby member is in ready mode. if not try giving the standby a reload.

If the standby member is now in a ready state you are now ready to do a live failover. Remember to enable the ports again on the switch side.

Conclusion

The process is not so bad as I thought. And there where no downtime involved. For me, I was missing ASDM and AnyConnect packages on the new standby node. I downloaded it from the existing primary node and then copied it to a USB disk. When the USB disk is plugged into the standby ASA I can then copy the files over the ASA flash.

From there on I could do a live failover and see the little “Active” light change on the physical ASA firewalls. With all traffic flowing uninterrupted. Mission accomplished.

macOS TFTP server

I have never really found a good TFTP for macOS. Is it funny that macOS is much used by network people but there isn’t a decent TFTP server?

Well, there is. macOS has it built-in, no GUI though. But that’s also fine, as long as you know to use it. It’s disabled by default, but you can start and stop it with the following commands.

The TFTP daemon uses the /private/tftpboot folder so we are going to copy the file there. Then set the correct permissions on the file.

There is a gotcha with the TFTP daemon, which is you cant copy a file to the TFTP daemon if that file does not already exist there.  To work around it you can just create a file and set the permission for it. Then your devices will just send data into the pre-created file.

Shrink VMDK disk

I have always thought that VMDK could only grow, so that has also been my default response to colleagues when they expanded a disk too much. Sure a storage vMotion could reclaim unused space in a thin disk, but the “down arrow” for storage capacity would never work. But then someone mentioned that he had done shrinking of disks a couple of times, I decided to investigate.

The official VMware kb isn’t too much help – somewhere discussing it on StackOverflow. But then I found an older post back from 2016 that seems to have found the approach so that’s what we are going to test out.

Disclaimer:

This is not supported in any way, use at your own responsibility. If you want a supported solution, then VMware converter in a v2v manner is kind of the only way. If you still want to try out the method, then be sure to have a valid backup! And by backup, it’s not a VMware snapshot.

Not supported:

From the VMware documentation, it seems shrinking disk is not allowed under the following circumstances:

  • The virtual machine is hosted on an ESX/ESXi server.ESX/ESXi Server can shrink the size of a virtual disk only when a virtual machine is exported. The space occupied by the virtual disk on the ESX/ESXi server, however, does not change.
  • The virtual machine has a Mac guest operating system.
  • You preallocated all the disk space to the virtual disk when you created it.
  • The virtual machine contains a snapshot.
  • The virtual machine is a linked clone or the parent of a linked clone.
  • The virtual disk is an independent disk in nonpersistent mode.
  • The file system is a journaling file system, such as an ext4, xfs, or jfs file system.

The test scenario:

I have a windows 2019 VM, here is the process I want to try out

  1. Expand VMDK disk in vCenter
  2. Extent disk in VM guest using diskpart
  3. Shrink disk in VM guest using diskpart
  4. calculate new sector size
  5. edit VM *.vmdk with the newly calculated sector size
  6. Storage migrate to other datastore
  7. Check if VM is still ok.

Walkthrough:

We start off with the VM. Its Windows 2019, original size is 40GB.
Disk is now extended with 5gb.
With a view from the esxi we can see the disk is also showing 45GB.
inside “win2019.vmdk” we can see the “extent description”. This is the number we have to change after the guest os filesystem has been shrunk.
Here we see the disk has been extended to 45GB and then shrunk down with 10GB.

Calculating the “extent description”:

So there is now 10GB free space we can shrink the VMDK with.

A virtual disk described as monolithic and flat consists of two files. One file contains the descriptor. The other file is the extent used to store virtual machine data.

Considering our existing extent
RW 94371840 VMFS “win2019-flat.vmdk”
This means that the file win2019-flat.vmdk is 94371840 sectors × 512 bytes/sector = 48318382080 bytes = 48318MB in size.

Let’s calculate the new value from GB to sectors.

36GB x 1024(mb) x 1024(kb) x 1024(byte) / 512byte pr sector = 75.497.472

before proceeding, we need to power off the VM. The .vmdk file is loaded into memory, so even if we can edit it now and start storage vMotion our changed value will just change back.

Letting vMotion do its magic
And after the boot of VM the disk is now shrunk. And we still have a working guest os.

Conclusion

It worked, we were able to add more space to the VM, extent, and shrink the guest os filesystem. We then calculated the number of sectors for the .vmdk file and storage vMotion did its magic and made the VMDK smaller in physical size.

I have also tried this in a couple of cases, also real life senairoes where people have added 4TB to much…. Then its sometimes easier to shrink than having to move files around.

VCD – Find free external IPs

Finding free public IPs in Cloud Director backed by NSX-V is not as easy as it should be. Some people will tell you to ping the scope and see what’s responding. But pinging is not reliable was of finding free IPs. Not every device is responding to ICMP messages.

Somewhere along the line, I found a guy on the VMware forum posting a script for finding available IPs in Cloud Director using the PowerCLI module for querying VCD and getting back IPs that are not allocated by an Edge. I have been using the script quite a bit since. His blog is not available today, but the code is still on the forum.

Now it’s also available here on the site with a bit more explanation on how to connect and use the function. I have been using it with NSX-V as backend, haven’t tried it with NSX-T at the network backend yet.

You can help yourself by copy and pasting the code snip into either PowerShell ISE or VisualCode. And since you need to install a cmdlet you need to run it with elevated rights. If you get a red message with importing the module it’s probably because of execution rights, you then need to run to command beneath. This is for allowing remote signed cmdlets to be executed.

Getting the names of the external networks with “get-externalnetwork”
Using the function to find available IPs in the selected external network

Change Veeam job repository

Having to move jobs to another repository is something that can be time-consuming. If you have to do it with the existing job you will need to disable the job, move the data to the new location, point the job to the new location and then enable the job again.

I found the other approach to creating new jobs easier. You can do it in the GUI, but when having 200+ jobs it could take some time. Instead, I did a small script to list all jobs located on the old repo.

The script also looks a the retention point and writes into the old job that is can be deleted after x days. If you have very long retention then this way is not feasible and you will probably have to move data and point exing job to the new location.

Hope somebody else can use it 🙂

NSX API – DLR L2 bridging

Here is a script for mass DLR L2 bridge creation. I had to bridge a couple of hundred VLAN to VXLAN, and while it was maybe faster to create it by hand I would not have learned anything.

The script is reading from a CSV file where I have all my info. Then loops through the entries and create a distributed port group and then initiates an L2 bridge. The VXLAN had been created post to this operation.

VCD – remove org and its items

Having a nice VRO job to create the VCD tenants with its VDCS, Edges and networks are nice. When having to clean up after testing its a pain to click through the GUI to first remote networks, then edges, then disable VDC, delete it, and final delete the org. A bit of PowerShell fu can help with the task, this is a quick and dirty script set of commands, but it works as intended.

Unlock ESXi root account

I often see people, including myself, lock themselves out of the ESXi web-based host client. It only locks you out from ssh and the web console. Password lockout is NOT active on the console/DCUI. Below is how you reset the counter and regain access.

Procedure to unlock the ESXi root

First, you need to gain ILO/IMM/IPMI or physical access to the server.

  1. At the console, press ALT+F1 to get to the ESXi shell. If a login shows up continue with step 3, otherwise continue with step 2. Change back to the login screen with ALT+F2.
  2. Login to the DCUI (to enable the ESXi Shell if not already done)
    1. Login with root and the correct password.
    2. Go to Troubleshooting Options
    3. Select Enable ESXi Shell
    4. Press CTRL+ALT+F1
  3. At the ESXi shell login with root and the password
  4. Run the following command to unlock the root account:

Veeam repository with local account

For the matter of security, I consider it a good idea to isolate the Veeam repository server from Active Directory. So that a compromised domain admin account or other can not gain access to the repository.

But when wanting to do add the repository to the VBR its failing and saying “Access Denied”.

Alright, a bit of googling and found a short and precise article from another guy having solved this problem.

What was the solution?

Open regedit on the repository server and navigate to following

Here you add a DWORD with the name of “LocalAccountTokenFilterPolicy” and value of “1”. This fixes the problem and without rebooting.

Now you can add the repository server to the VBR. I always forgot where to find the info for the reg hack, so now it’s here so future Jesper can find it 🙂

Using dd – disk cloning

dd (disk duplication) is a utility that can read raw data of a disk, even if the Mac doesn’t understand the filesystem.

I have used it before in p2v a physical to a virtual server. For details take a look at this article.

  • if= specifies input path (file, or device)
  • of= specifies output path (file, or device)
  • bs=n sets both input and output block size (optional, default=512 byte blocks)
  • conv=noerror,sync tells dd to be fault-tolerant and ignore read errors (optional)

If the operation stops with an I/O error, trying to salvage all readable data with conv=noerror,sync.
This option can often recover a dead hard drive or an unreadable file, but it does not repair the error.

Make a clone of a disk:

In this case, I was fooling around with a Microsoft StorSimple appliance and wanted to have a backup of it to go back to if I messed it up too badly. This is not the first time I have done it, and sure not the last time, and always forgot the commands, so to future Jesper, here goes.

Open Terminal and try this:

  1. Attach and identify the source disk:

2. if mounted, unmount the source disk:

Copy the source disk to the desktop:

If you want to save space, make it gzip it with:

Failed disk backup:

If the copy fails due to disk errors, try using the following command at step 3:

dd will take much longer using this option, errors are written as null bytes. fsck / chkdsk the disk afterward.

Notes:

  • Using /dev/rdisk# (raw disk) in place of /dev/disk# is much faster
  • You can check progress, while the command is running, by pressing Ctrl-T
  • If you can attach both disks, copy directly: sudo dd if=/dev/disk2 of=/dev/disk3 bs=64k

VHDX Native boot

I stumbled upon the concept of VHDX native boot. Its a rather old feature but very overlooked. When I had a windows laptop I would have loved this feature. Being able to multiboot so you could format your PC with still have the possibility to boot the old installation.

Its fairly simple, you OS is contained within a VHDX on your disk. The boot loader on the disk then has an entry of that VHDX file. Simple but yet powerful. You could have another VHDX for all your data and then a VHDX for OS. Then when booting into each of your environments you would have your data with you.

How to:

If you have a native os install today you can still use this feature. So it’s easy to convert into VHDX native boot. First, we need to do a bit of diskpart, then use dism to install the OS into the VHDX. But this can all be done while your old OS is running and you don’t have to prep USB keys etc.

Diskpart

We will create a new VHDX file with parameters of size and type.

DISM:

Now we have a VHDX file that Is attached, formated, and has a drive letter. On to install of your OS. You will need your install media, for Windows 10 you normally use index 1, if you are installing Windows server index 1 is probably the server core install if you want GUI then install with index 2.

Afterward, bcdboot creates the boot entry into the boot loader and lastly, you can change the description on the entry in the boot loader, so that you can remember what has been installed.

Conclusion:

You can now reboot and have the choice to boot your newly created VHDX with your fresh installed OS. So your company installed Windows laptop without admin rights, you can now boot into your private install so you have a company and private side of your work PC.

StorSimple 8100 – reuse

I bought a Microsoft StorSimple 8100 unit. The only catch, it did not contain its SSDs and the password for it was unknown. Fair enough.

Its quite an interesting unit, hardware-wise. Its a 2U with 2x750W PSU, 12×3,5 SAS bays, and 2x compute nodes. Each compute node is in fact a Xeratex CS-6000-AB containing:

  • 1* E5-2648L 1,8Ghz 8 core 70Watt CPU
  • 4* 8GB DDR3 memory
  • 1* LSI/Avago/Broadcom 2308 SAS HBA (1*SFF8088 and 1*internal link)
  • 1* Mellanonx ConnectX3 10/40/56Gbit dual QSFP
  • 1* 128GB SSD for OS.

The two compute nodes share a Xeratex HB1235 enclosure with the 12 3,5″ drive bays. This enclosure is used for many other storage vendors as HPE 3PAR or Dell Compellent SANs.

IPMI/BMC enable

Not having a DisplayPort to connect a screen so you can see what is going on is making this a very proprietary piece of hardware. But when having access to the IPMI then all of sudden it becomes easy to reuse the hardware for something different than the StorSimple software.

This is how to enable the IPMI/BMC hardware.

  1. Reseat one of the controllers or power cycle the appliance with a console cable connected
  2. Press Esc to enter the boot options
  3. Select “Setup Utility” from the list
  4. It will prompt for a password (E1aD8wAbMxB3XcpjwVKD)
  5. Go to Advanced tab
  6. Go to  IPMI BMC Configuration
  7. Go to BMC Configuration
  8. Scroll down till you get to the bottom and you will see the network configuration
  9. Select LAN Channel number 1 and static IP source
  10. Enter the IP, subnet, and gateway
  11. Press F10 to save and exit
  12. Log into the BMC with web browser and access the console from there Log in Username: admin  Password: admin
The IPMI/BMC interface of Xeratex CS-6000 node

Now you can open a java based KVM tool to get the display from the node and do what you want. Awesome!

Java….

But there is a small catch, you can’t just open and run the IPMI. The firmware is old and uses encryption algorithms that are not allowed anymore. So you need to change the security properties of your java install and run the IPMI in an Internet Explorer running compatibility mode.

This is quite an easy fix. What I did was to open notepad as administrator, and edit the following file:

find and comment out the line that starts with “jdk.jar.disabledAlgorithms” by prefixing a #. Note that this will allow jar files signed with any algorithms to run, which can is to be considered insecure! But for us a necessary measure for getting access to the IPMI.

StorSimple software

Each compute node is using VHDX native-boot. So the SSD has a boot loader, and then each VHDX is in that boot loader. That means that they can deploy a newer version or factory reset by switching over to another VHDX disk. I was actually not aware of something like VHDX native boot, but its a very nice feature. For sure going to use that on my windows based laptop in the future. So much easier than having to do the native OS install.

The StorSimple software is based on Windows Server 2012R2. You are normally only able to use use the console connection for direct management, but it actually also has an IPMI/BMC feature on each compute node you can look deeper into the system.

Since I did not have the device password the StorSimple software could do nothing. So I got my fingers on PCUnlocker, a password reset tool. Booted through IPMI, where I could attach the VHDX file and have it reset the passwords of the administrator. This account was also disabled, but PCUnlocker did also take care of that part.

Now boot back into the StorSimple software I could now choose an administrator account, type in my new password and now I had access to a cmd. It was using server core install, so no GUI but that’s ok because now I had access to all the other HCS PowerShell cmdlets.

Unfortunately the former owner had also tried to mess around with it, so the factory default VHDX images and the compute node signatures did not match and therefore the “reset-hcsfactorydefault” could not validate the factory default images. Bummer.

Many of the HCS cmdlets where PowerShell cmdlets referring to a DDL, so no way to see what was going on. But the “test-hcsfactoryimage” and reset/initialize scripts where full-blown PowerShell. So from there, I could see what was checked for the VHDX image to validate. I actually did a bypass on the validation, and did the reset command, but after each node had generated a new VHDX from the factory VHDX files I booted but was stuck in the boot state of HCS software.

I found an eagerness to find a way to fix it, but then again the time spent would not payout. You need an Azure subscription to actually manage StorSimple since there is no local GUI, only the serial console. So I decided to install Windows Server 2019 in it instead. 🙂

Conclusion

It’s a nice piece of hardware, StorSimple should have been nice to use if it was not depended on Azure. I now have a 2-node possibility to run an HCI cluster running Storage Spaces and with a failover cluster, presented to each node with CSV volumes. I could run HyperV and have a 2U box with full redundancy. I still feel the eager to fix the StorSimple software but not for now 🙂

FreeBSD – install phpipam

Quick guide on how to install phpipam on FreeBSD. I will assume that you know how to install FreeBSD 🙂

Remember to have a freshly updated server and use sudo instead of directly root access.

Configure mysql

We won’t do any tuning to mysql, just create a user and database and lets go.

Configure phpipam

Get phpipam and put in www dir. Use git to get code, this will also make it easier for version updates later on.

Updating phpipam

Finish up by opening the web interface and follow upgrade procedure.

Configure nginx

Make nginx start on boot and backup the original config. We will then add our own.

After we now have the backup, lets add the content beneath to nginx.conf.

Configure PHP/FPM

Lets make a production ini file and afterwards setup php-fpm config file.

Open the file /usr/local/etc/php-fpm.d/www.conf and uncomment the following lines.

Conclusion

We have now installed all the required components, you should now reboot the server and check if all the services is coming up automatically. If so you can proceed and access the web interface of your new phpipam installation. Then follow the guide on how to get setup.

NetApp – ServiceProcessor stuck updating

After update to OnTap 9.7 the service processors where stuck in “updating”. They never came up again, not even after rebooting it.

Procedure:

  • Disable auto update
  • Reboot one SP, wait for it to show online.
  • Run the update parameter manuel
  • If its online and updated then enable auto update again.
Here we see that the ctrl02 is online again, but with wrong firmware.
After Manuel update we now have the correct firmware and they are online. Ready to enable autoupdate

https://kb.netapp.com/app/answers/answer_view/a_id/1028746/~/service-processor-firmware-update-fails-

Cloud Director 10.1 released

Been using vCloud since version 5.1. After a brief love affair with something called “Azure Pack” we put all our focus into vCloud.

8.20 was the first sign of heartbeats coming from VCD. We got confirmation that vCloud was for sure the platform that we were and had been looking for. Now we see the 10.1 released and from my point of view it’s a big one, may things change in GUI as in infrastructure. This release is also the final farewell to the old flex GUI.

First off we have to address the naming, I always liked the vCloud term, for me a strong brand. So a bit sad to see that go and now we have to get used to the Cloud Director instead. Thankfully we can still use the acronym VCD for VMware Cloud Director. #LongLiveVCD.

In the next few points, I will address some of the major things within this release.

APIs

We use a lot of the functionality of the APIs of VCD. Since we see that the development of VCD is changing into higher gear, so is the deprecation of the older API versions. For a small service provider, it’s always hard to revisit automation already working with existing APIs. When going on board 10.1 we have to go through a couple of workflows to update the to use the new 34.0 API. But on the other side, it’s also a good chance to refactor and optimize.

  • VMware Cloud Director API version 29 and below are not supported.
  • VMware Cloud Director API version 30.0 is deprecated and will become unsupported after VMware Cloud Director 10.1
  • VMware Cloud Director API version 31.0 is deprecated.

NSX-T feature improvements

More of the core NSX-T features is now available through VCD.

  • IPSec VPN
  • Dedicated External Network
  • BGP and Route Advertisement

We have been looking from the side for NSX-T development to reach an acceptable level for some time. NSX-V is still doing a good job. As someone who right now is standing up a new 16 node VMware cluster as a new provider VDC, I would have wished for it to be 6 months later so that all NSX-T functionality was ready and we could hopefully solo use NSX-T.

But we have to look into maybe having two 8 node clusters for NSX-V and on for NSX-T so we can already now start to transition to NSX-T…

But the good thing about being a VMware customer is that you are not left in the dust. There have been already been created migration tools for NSX-V > NSX-T, NSX-T Data Center Migration Coordinator, but it had no integration to VCD. which bring me to the next point!

NSX-V to NSX-T VCD Migration Tool

This is a way of helping us transition from NSX-V to NSX-T as we are seeing NSX-V lacking to the end of support in January 2021.

Before we could still do a new provider VDC that was backed by NSX-T controller and then start to move workloads over to the new cluster and at the time had to use NSX-T functionality, but all in a manual process.

There is now an automated way to do it, which is VCD aware. The approach will require a new cluster since NSX-V and NSX-T can’t coexist in the same cluster. From the Whats New in 10.1 it stats that the workflow will help with following

  • Automates migration of vCD metadata and workloads from NSX-V to NSX-T
  • Migrate per Org VDC migration to reduce maintenance window to single tenant
  • Minimize network downtime with bridged networks during migration
  • Live migrate with vMotion to ensure non-disruption to user workloads
  • Keep source VDC configuration and environment as-is to allow rollback
Before live migration
After live migration

Tomas did a good discussion on this subject

SSL and Certificate Management

This seems like something to read up on carefully. In short, VCD does not trust endpoint certificates unless they have been imported to the trust store.

There is a tool helping with the import, trust-infra-certs, that automatically connect to the endpoint, grabbing and importing the certificate. If this is not done successfully you will not be able to talk to those endpoints after upgrading to VCD 10.1.

App Launchpad

A new feature to help introduce a marketplace with the help of the content from Bitnami. From there we can now offer customers to easily find, deploy and manage new workloads. Not just as VMs but also as containers.

Daniel did an excellent write up on this subject.

Conclusion

There is still a lot more in this release to talk about, CSE2.6, OSE1.5, Terraform 2.7 provider, etc. read more from the official release notes.

Might have had to write a disclaimer for the length of this post and the lack of interesting pictures, will try to improve for next time.

I love to see VCD take flight. We are looking forward being part of the future journey where things like Bitnami and App Launchpad together with more NSX-T functionality and a whole lot of other features helps us Cloud Providers to help other business to there digital transformation .

Big shout out to VMware and the VCD team!

vCloud SAML authentication – Automation

Cant say that I did everything by my self in this post, I had a great great help from my college and friend Kasper Hansen. Also gotten a great help from the vExpert community, especially Tom Fojta.

In my last post I found out how to setup vCloud SAML against AzureAD. Now we are gonna look on how to automate each tenant to use the same AzureAD. In these days everybody have either a Microsoft og AzureAD account, so this way its easy to invite them as guest users and this way have controlled access but also ensure that vCloud users have MFA enabled.

We use VRO for the creation of vCloud tenants, in this flow we are now going to introduce a new workflow that will do following. Although the workflow is just a restcall to trigger an event in Azure Automation.

  • Create AzureAD Groups for admin and viewer
  • Post federation metadata to vCloud tenant
  • Post federation groups to vCloud tenant

Enable SAML

Fojta have some very good articles on his blog on the basic setup of SAML to different IDP systems. I also did a piece on it where Azure where the IDP provider.

Because we want to have all organisations linked to the same SAML app in Azure we need to have the same SAML certificate on all organisations. You can only do this with the API, but what the documentation did not say was that the certificate needs to be trusted by the keystore, the java keystore of the cells.

Create a self-signed certificate and make vCloud trust it

These commands will help you create a certificate and a private key in the needed pkcs8 format and certificate in the x509 format.

When you have done the new self-signed certificate you need to import it to each and one of your cells. After import you will need to restart the cells. One of the errors I did here way that I tried to import a .pem where the private key and certificate where combines, that won’t work. Only import the certificate.

Publish federation settings to API

We where having a lot of trial and error in this step, because that vCloud did not trust the certificate. Each time a put where done to the API the log complained. /opt/vmware/vcloud-director/logs/vcloud-container-debug.log it showed “Failed to generate keystore | requestId=<id>,request=PUT.”

Following is a example done in PowerShell, insert your self-signed certificate where it says —–END/BEGIN CERTIFICATE—–

Publish SAML groups to API

Now that the SAML metadata/certificate is uploaded and in place we need to add groups to tenant. You can read more about what groups/users should be imported in my other SAML blog post.

Each tenant have its own role ids, so when doing automation with group import we need to query the vCloud API and get the role ids. There is a specific query API to get data. When using a system account we need to specify a “VCLOUD-TENANT-CONTEXT” in the header of the request. This we we can query a tenant context from a system account.

After we got the role id we can now send up the group together with the role id and this was be able to authenticate based on a SAML group from AzureAD.

Conclusion

Now we have all the pieces for making automation where we can enable a tenant for SAML authentication and afterwards import f.eks. a viewer and admin group. External users will then be invited to the AzureAD, imported into the right group and now they have access to the their tenant. We can help the organisation secure the access to their virtual datacenter with MFA and they will have single sign-on with there own user that originates from their own AzureAD or Microsoft account.

A service library will be made where users can be invited to the tenant organisation. So that when one user have been invited that user will be able to invite its colleagues.

Homelab – v1

This will be a short blog series of the new setup and how you can start to do your own homelab.

The basic idea of a homelab

I have always had a homelab, small, but enough to learn and the more you learn the bigger your need is. The first homelab consisted of 2* Apple Mac mini. The Apple Mac mini is very power efficient and very quiet. Not the beefiest hardware, but just enough to be able to run a vCenter and have vSAN running.

  • Apple Mac Mini v5.1 mid-2011 A1347
  • 2.3 GHz Core i5 (I5-2415M)
  • 16GB DDR3 memory
  • Dual Drive kit
  • 256GB Cache disk
  • 600GB capacity disk

They were mounted in a Sonnet MacRack mini 1U enclosure. Which have been perfect for many years. In my small setup I have been running my pFsense firewall and all sorts of small VMs, due to the small memory amount I was primary FreeBSD VMs with services as Zabbix, Weewx, OpenHAB, Unify controller, TOR and things like that. All stuff to play around with besides VMware of cause.

“Homelabbing” is where is see people learn and are having fun, without breaking too much.

The idea of new homelab

I have always had a way higher power bill than other “normal” people”. Servers, NAS and home automation gear standing around are not good for you power bill. And that’s also why my first homelab was made of Mac mini.

So instead of having huge servers in the garage or basement, I have always tried to keep the footprint down. The WOF(WifeAprovalFactor) also makes a hit here 🙂

I have a wall-mounted 19″ rack with 12U and 600mm depth. Placed in the garage where noise is not a problem anymore.

I want to run an all-flash VMware vSAN cluster with three nodes. I don’t want only two hosts and a witness appliance, even if it works and it is a fully supported concept for small- or branch offices. I want a real scale. Each server should have one cache device and at least one SSD for the capacity tier. I went all-in and decided to go with two SSDs for capacity. All servers have to be connected with 10Gbit SFP+ for vSAN and vMotion.

Conclusion of upcoming homelab

  • Small footprint, both power, and space.
  • 3 node all-flash vSAN cluster
  • 10Gbit SFP+ networking
  • Formfactor must be rack

The new hardware

Decided to go with Supermicro hardware. They have IPMI and actually some of the E300 series is now on the VMware compatibility list.

Supermicro kits such as the E300 are a very popular choices amongst the VMware community. It got a powerful Xeon-based CPUs and support for up to 128GB of memory, it is perfect for running a killer vSphere/vSAN setup and still keeping cost, noise and power bill down.

BOM

Here is a list of what the hardware consists of. This gives a hell of horsepower for a homelab and plenty of memory and CPU for doing nested environment so test our NSX-V to NSX-T migrations etc.

NumberItem
3Supermicro E300-8D 4core Intel D-1518 2Ghz
3Intel 790p 128GB disk for cache tier
3Supermicro Riser card
3Supermicro SATA-DOM 32GB
12Samsung 32GB DDR4 memory modules
3Intel 600GB SSD for capacity tier
3Supermicro Rackmounts

Now the hardware is documented. Next will be rack and stack and how the environment will be designed 🙂

vCloud SAML authentication

vCloud have LDAP, SAML and local users as an option for tenant authentication. In this post, we are looking into SAML integration. With AzureAD.

The cool thing about AzureAD is that you will gain the MFA option out of the box, and when tenants want access we can also invite them from their own AzureAD tenant into the resource AzureAD tenant. This gives flexibility and overview of who has access.

ADFS is also an option, but there you need to keep your own infrastructure with a resource AD/ADFS and furthermore need a 3. party MFA solution.

Process:

  • Setup Enterprise app in desired resource AzureAD
  • Setup claims
  • Set federation entity id for tenant
  • Import vCloud federation metadata to AzureAD
  • Import AzureAD enterprise app federation metadata to vCloud
  • Setup allowed users/groups in vCloud

AzureAD

Let’s get started with Azure AD configuration. Login to your AzureAD portal https://portal.azure.com. Navigate to “Azure Active Directory” > “Enterprise App” and press “New Application”. Choose “Non-gallery application”. Give it the name “vCloud SAML test” and press “Add”. This will take a couple of minutes.

Navigate back to “Enterprise Apps” > “All applications” and choose your newly created App.

For test purpose, add/assign a test user to the app. This is under “Users and groups”. This user will be able to login to the enterprise app with AzureAD.

Now go to “Single Sign-on”. This will now ask for the sign-on method, and here we will choose “SAML”. This will then take us to the SAML setup. The first thing to do is importing the metadata from the cloud.

You will find the metadata by logging in to vCloud, go to the tenant, under “administration” > “federation” tab. Enter the URL for the tenant as a entity id, apply and afterwards download the metadata from the link.

You will find the metadata by logging in to vCloud, go to the tenant, under administration choose the federation tab. Enter the URL for the tenant as a entity id, apply and afterwards download the metadata from the link.

In azureAD “Upload Metadata” and chose the downloaded file from vCloud. This will give AzureAD the knowledge of where to redirect and accept request from.

vCloud can validate a couple of user/group parameters. Vmware documentation. So we will add some claims to Azure AD.

Now we will need to download the AzureAD metadata and import into vCloud. Fetch the data by pressing “Download” to the “Federation Metadata XML”.

Head over to vCloud tenant federation page again. Paste the content from the download metadata file. check the “Use SAML identity” and apply. Now we are almost ready to try it out. But first, head over to “Users” tab in vCloud. We need to add the user/role to whom are allowed to gain vCloud Access.

Here we put in the mail address and role of the user from Azure AD. When the SAML response then returns to vCloud then vCloud can see it been authenticated in Azure AD and that the user is an Org admin.

Next step would be to use groups and roles so that we can put users into groups in Azure AD and that way manage access for the tenant. But after this, we can now head to the tenant URL. We will then be redirected to the Azure AD login page, login and accept to MFA so that we can be redirected to our vCloud tenant.

And voila, we have logged into our vCloud tenant with Azure AD.

Troubleshooting:

When I first started this project I was using a GUID as a vCloud entity id. That meant that I could get it to work with ADFS but not AzureAD. I went full mole on the troubleshooting.

In the end, I intercepted the SAML responses. These are encoded in base64, easy task to decode. And afterwards, I got the XML that either ADFS or AzureAD is sending back. I could then compare them, and I saw som <ds> tags to the cert that wasn’t on in the response from AzureAD. Unfortunately, that was a duck and meant nothing.

By tailing the log from vCloud, tail -f /opt/vmware/vcloud-director/logs/vcloud-container-debug.log, I could get some hints when the SAML auth failed.

org.opensaml.common.SAMLException: Local entity is not the intended audience of the assertion in at least one AudienceRestriction

doing a bit more googling and found out that I should be looking at the <audience> tag from the two SAML responses. And yes, that made some sense.

Azure AD sets the value of this element to the value of Issuer element of the AuthnRequest that initiated the sign-on. To evaluate the Audience value, use the value of the App ID URI that was specified during application registration.
Like the Issuer value, the Audience value must exactly match one of the service principal names that represents the cloud service in Azure AD. However, if the value of the Issuer element is not a URI value, the Audience value in the response is the Issuer value prefixed with spn:.

https://stackoverflow.com/questions/38978298/azuread-jwt-token-audience-claim-prefix-makes-jwt-token-invalid

And that was the problem, spn: prefix when not using a URL as entity id. Changing it to the URL made it work.

Maybe this is obvious to the world, but I didn’t know it, but glad my troubleshooting skills where sufficient 🙂