May 062018

I’m a big fan of MFA, specifically Duo Security‘s product (I did a corporate blog post here). I’ve been using this product for some time and use it for an extra level of protection on my workstations, servers, and customer sites. I liked it so much so that my company (Digitally Accurate Inc.) became a partner and now resells the services.

Today I want to write about a couple issues I had when deploying the pam_duo module on CentOS Linux 7. The original duo guide can be found at, however while it did work for the most part, I noticed there were some issues with the pam configuration files, especially if you are wanting to use Duo MFA with usernames and passwords, and not keys for authentication.

A symptom of the issue: I noticed that when following the instructions on the website for deployment, after entering the username, it would skip the password prompt, and go right for DUO authentication, completely bypassing the password all together. I’m assuming this is because the guide was written for key authentication, but I figured I’d do a quick crash-course post on the topic and create a simple guide. I also noticed that sometimes even if an incorrect password was typed in, it would allow authentication if DUO passed as successful.

Ultimately I decided to learn about PAM, understand what it was doing, and finally configure it properly. Using the guide below I can confirm the password and MFA authentication operate correctly.

To configure Duo MFA on CentOS 7 for use with usernames and passwords

First and foremost, you must log in to your Duo Account and go to applications, click “Protect an Application” and select “Unix Application”. Configure the application and document/log your ikey, secret key, and API hostname.

Now we want to create a yum repo where we can install, and keep the pam_duo module up to date. Create the file /etc/yum.repos.d/duosec.repo and then populate it with the following:

name=Duo Security Repository

We’ll need to install the signging key that the repo uses, and then install the duo_unix package. By using yum, we’ll be able to keep this package regularly up to date when we update the server. Run the following commands:

rpm --import
yum install duo_unix

Configure the pam_duo module by editing the /etc/duo/pam_duo.conf file. You’ll need to populate the lines with your ikey, secret key, and API hostname that you documented above. We use “failmode=safe” so that in the event of an internet disconnection, we can still login to the server without duo. It’s safe to enable this fail-safe, as the purpose is to protect it against the internet. Please see below:

; Duo integration key
; Duo secret key
; Duo API host
host =
; Send command for Duo Push authentication
pushinfo = yes
; failmode safe if no internet it works (secure locks it up)
failmode = safe

Configure sshd to allow Challenge Response Authentication by editing /etc/ssh/sshd_config, then locate and change “ChallengeResponseAuthentication” to yes. Please note that the line should already be there, and you should simply have to move the comment symbol to comment the old line, and uncomment the below line as shown below:

ChallengeResponseAuthentication yes

And now it gets tricky… We need to edit the pam authentication files to incorporate the Duo MFA service in to it’s authentication process. I highly recommend that throughout this, you open (and leave open) an additional SSH session, so that if you make a change in error and lock yourself out, you can use the extra SSH session to revert any changes to the system to re-allow access. It’s always best to make a backup and copy of these files so you can easily revert if needed.

DISCLAIMER: I am not responsable if you lock yourself out of your system. Please make sure that you have an extra SSH session open so that you can revert changes. It is assumed you are aware of the seriousness of the changes you are making and that you are taking all precautions (including a backup) to protect yourself from any errors.

Essentially two files are used for authentication that we need to modify. One file is for SSH logins, and the other is for console logins. In my case, I wanted to protect both methods. You can do either, or both. If you are doing both, it may be a good idea to test with SSH, before making modifications to your console login, to make sure your settings are correct. Please see below for the modifications to enable pam_duo:

/etc/pam.d/password-auth (this file is used for SSH authentication)

# This file is auto-generated.
# User changes will be destroyed the next time authconfig is run.
auth        required
auth        required delay=2000000
#auth        sufficient nullok try_first_pass
auth        requisite nullok try_first_pass
auth  sufficient
auth        requisite uid >= 1000 quiet_success
auth        required

account     required
account     sufficient
account     sufficient uid < 1000 quiet
account     required

password    requisite try_first_pass local_users_only retry=3 authtok_type=
password    sufficient sha512 shadow nullok try_first_pass use_authtok

password    required

session     optional revoke
session     required
-session     optional
session     [success=1 default=ignore] service in crond quiet use_uid
session     required

/etc/pam.d/system-auth (this file is used for console authentication)

auth        required
auth        sufficient
#auth        sufficient nullok try_first_pass
# Next two lines are for DUO mod
auth        requisite nullok try_first_pass
auth        sufficient
auth        requisite uid >= 1000 quiet_success
auth        required

account     required
account     sufficient
account     sufficient uid < 1000 quiet
account     required

password    requisite try_first_pass local_users_only retry=3 authtok_type= ucredit=-1 lcredit=-1 dcredit=-1 ocredit=-1
password    sufficient sha512 shadow nullok try_first_pass use_authtok remember=5
password    required

session     optional revoke
session     required
-session     optional
session     [success=1 default=ignore] service in crond quiet use_uid
session     required

Now, we must restart sshd for the changes to take affect. Please make sure you have your extra SSH session open in the event you need to rollback your /etc/pam.d/ files. Restart the sshd service using the following command:

service sshd restart

Attempt to open a new SSH session to your server. It should now ask for a username, password, and then prompt for Duo authentication. And you’re done!

More information on Duo Multi Factor Authentication (MFA) can be found here.

May 012018
Fedora Logo

When attempting to upgrade from Fedora Core 27 to Fedora Core 28, it may fail on the nss-pem package.


I spent some time trying to find the solution for this, and came across numerous posts on the “Red Hat Bugzilla”, particularly this post.

See below example:

[root@SYSTEMZ01 ~]# dnf system-upgrade download --releasever=28
Before you continue ensure that your system is fully upgraded by running "dnf --refresh upgrade". Do you want to continue [y/N]: y
Fedora 28 - x86_64 - Updates                                                                                                                                $
Fedora 28 - x86_64                                                                                                                                          $
google-chrome                                                                                                                                               $
RPM Fusion for Fedora 28 - Free - Updates                                                                                                                   $
RPM Fusion for Fedora 28 - Free                                                                                                                             $
RPM Fusion for Fedora 28 - Nonfree - Updates                                                                                                                $
RPM Fusion for Fedora 28 - Nonfree                                                                                                                          $
skype (stable)                                                                                                                                              $
Last metadata expiration check: 0:00:00 ago on Tue 01 May 2018 04:28:04 PM MDT.
 Problem: nss-pem-1.0.3-6.fc27.i686 has inferior architecture
  - nss-pem-1.0.3-6.fc27.x86_64 does not belong to a distupgrade repository
  - problem with installed package nss-pem-1.0.3-6.fc27.i686

To resolve this, manually install the nss-pem packages from FC28 prior to the upgrade using the following command.

dnf install nss-pem-1.0.3-9.fc28 --releasever=28

After doing so, re-attempt to upgrade and the upgrade should now proceed.

Jan 182018

The Problem

I run a Sophos UTM firewall appliance in my VMware vSphere environment and noticed the other day that I was getting warnings on the space used on the ESXi host for the thin-provisioned vmdk file for the guest VM. I thought “Hey, this is weird”, so I enabled SSH and logged in to check my volumes. Everything looked fine and my disk usage was great! So what gives?

After spending some more time troubleshooting and not finding much, I thought to myself “What if it’s not unmapping unused blocks from the vmdk to the host ESXi machine?”. What is unampping you ask? When files get deleted in a guest VM, the free blocks aren’t automatically “unmapped” and released back to the host hypervisor in some cases.

Two things need to happen:

  1. The guest VM has to release these blocks (notify the hypervisor that it’s not using them, making the vmdk smaller)
  2. The host has to reclaim these and issue the unmap command to the storage (freeing up the space on the SAN/storage itself)

On a side note: In ESXi 6.5 and when using VMFS version 6 (VMFS6), the datastores can be configured for automatic unmapping. You can still kick it off manually, but many administrators would prefer it to happen automatically in the background with low priority (low I/O).

Most of my guest VMs automatically do the first step (releasing the blocks back to the host). On Windows this occurs with the defrag utility which issues trim commands and “trims” the volumes. On linux this occurs with the fstrim command. All my guest VMs do this automatically with the exception being the Sophos UTM appliance.

The fix

First, a warning: Enable SSH on the Sophos UTM at your own risk. You need to know what you are doing, this also may pose a security risk and should be disabled once your are finished. You’ll need to “su” to root once you log in with the “loginuser” account.

This fix not only applies to the Sophos UTM, but most other linux based guest virtual machines.

Now to fix the issue, I used the “df” command which provides a list of the filesystems, their mount points, and storage free for those fileystems. I’ve included an example below (this wasn’t the full list):

hostname:/root # df
Filesystem                       1K-blocks     Used Available Use% Mounted on
/dev/sda6                          5412452  2832960   2281512  56% /
udev                               3059712       72   3059640   1% /dev
tmpfs                              3059712      100   3059612   1% /dev/shm
/dev/sda1                           338875    15755    301104   5% /boot
/dev/sda5                         98447760 13659464  79513880  15% /var/storage
/dev/sda7                        129002700  4624468 117474220   4% /var/log
/dev/sda8                          5284460   274992   4717988   6% /tmp
/dev                               3059712       72   3059640   1% /var/storage/chroot-clientlessvpn/dev

You’ll need to run the fstrim command on every mountpoint for file systems “/dev/sdaX” (X means you’ll be doing this for multiple mountpoints). In the example above, you’ll need to run it on “/”, “/boot”, “/var/storage”, “/var/log”, “/tmp”, and any other mountpoints that use “/dev/sdaX” filesytems.

Two examples:

fstrim / -v

fstrim / -v



fstrim /var/storage -v

fstrim /var/storage -v



Again, you’ll repeat this for all mount points for your /dev/sdaX storage (X is replaced with the volume number). The command above only works with mountpoints, and not the actual device mappings.

Time to release the unused blocks to the SAN:

The above completes the first step of releasing the storage back to the host. Now you can either let the automatic unmap occur slowly overtime if you’re using VMFS6, or you can manually kick it off. I decided to manually kick it off using the steps I have listed at:

You’ll need to use esxcli to do this. I simply enabled SSH on my ESXi hosts temporarily.

Please note: Using the unmap command on ESXi hosts is very storage I/O intensive. Do this during maintenance window, or at a time of low I/O as this will perform MAJOR I/O on your hosts…

I issue the command (replace “DATASTORENAME” with the name of your datastore):

esxcli storage vmfs unmap --volume-label=DATASTORENAME --reclaim-unit=8

This could run for hours, possibly days depending on your “reclaim-unit” size (this is the block size of the unit you’re trying to reclaim from the VMFS file-system). In this example I choose 8, but most people do something larger like 100, or 200 to reduce the load and time for the command to complete (lower values look for smaller chunks of free space, so the command takes longer to execute).

I let this run for 2 hours on a 10TB datastore, however it may take way longer (possibly 6+ hours, to days).

Finally, not only are we are left with a smaller vmdk file, but we’ve released the space back to the SAN as well!

Apr 122014

Recently I decided it was time to beef up my storage link between my demonstration vSphere environment and my storage system. My existing setup included a single HP DL360p Gen8, connected to a Synology DS1813+ via NFS.

I went out and purchased the appropriate (and compatible) HP 4 x 1Gb Server NIC (Broadcom based, 4 ports), and connected the Synology device directly to the new server NIC (all 4 ports). I went ahead and configured an iSCSI Target using a File LUN with ALUA (Advanced LUN features). Configured the NICs on both the vSphere side, and on the Synology side, and enabled Jumbo frames of 9000 bytes.

I connected to the iSCSI LUN, and created a VMFS volume. I then configured Round Robin MPIO on the vSphere side of things (as always I made sure to enable “Multiple iSCSI initators” on the Synology side).

I started to migrate some VMs over to the iSCSI LUN. At first I noticed it was going extremely slow. I confirmed that traffic was being passed across all NICs (also verified that all paths were active). After the migration completed I decided to shut down the VMs and restart to compare boot times. Booting from the iSCSI LUN was absolutely horrible, the VMs took forever to boot up. Keep in mind I’m very familiar with vSphere (my company is a VMWare partner), so I know how to properly configure Round Robin, iSCSI, and MPIO.

I then decided to tweak some settings on the ESXi side of things. I configured the Round Robin policy to IOPS=1, which helped a bit. Then changed the RR policy to bytes=8800 which after numerous other tweaks, I determined achieved the highest performance to the storage system using iSCSI.

This config was used for a couple weeks, but ultimately I was very unsatisfied with the performance. I know it’s not very accurate, but looking at the Synology resource monitor, each gigabit link over iSCSI was only achieving 10-15MB/sec under high load (single contiguous copies) that should have resulted in 100MB/sec and higher per link. The combined LAN throughput as reported by the Synology device across all 4 gigabit links never exceeded 80MB/sec. File transfers inside of the virtual machines couldn’t get higher then 20MB/sec.

I have a VMWare vDP (VMWare Data Protection) test VM configured, which includes a performance analyzer inside of the configuration interface. I decided to use this to test some specs (I’m too lazy to actually configure a real IO/throughput test since I know I won’t be continuing to use iSCSI on the Synology with the horrible performance I’m getting). The performance analyzer tests run for 30-60 minutes, and measure writes and reads in MB/sec, and Seeks in seconds. I tested 3 different datastores.


Synology  DS1813+ NFS over 1 X Gigabit link (1500MTU):

Read 81.2MB/sec, Write 79.8MB/sec, 961.6 Seeks/sec

Synology DS1813+ iSCSI over 4 x Gigabit links configured in MPIO Round Robin BYTES=8800 (9000MTU):

Read 36.9MB/sec, Write 41.1MB/sec, 399.0 Seeks/sec

Custom built 8 year old computer running Linux MD Raid 5 running NFS with 1 X Gigabit NIC (1500MTU):

Read 94.2MB/sec, Write 97.9MB/sec, 1431.7 Seeks/sec


Can someone say WTF?!?!?!?! As you can see, it appears there is a major performance hit with the DS1813+ using 4 Gigabit MPIO iSCSI with Round Robin. It’s half the speed of a single link 1 X Gigabit NFS connection. Keep in mind I purchased the extra memory module for my DS1813+ so it has 4GB of memory.

I’m kind of choked I spent the money on the extra server NIC (as it was over $500.00), I’m also surprised that my custom built NFS server from 8 years ago (drives are 4 years old) with 5 drives is performing better then my 8 drive DS1813+. All drives used in both the Synology and Custom built NFS box are Seagate Barracuda 7200RPM drives (Custom box has 5 X 1TB drives configured RAID5, the Synology has 8 x 3TB drives configured in RAID 5).

I won’t be using iSCSI  or iSCSI MPIO again with the DS1813+ and actually plan on retiring it as my main datastore for vSphere. I’ve finally decided to bite the bullet and purchase an HP MSA2024 (Dual Controller with 4 X 10Gb SFP+ ports) to provide storage for my vSphere test/demo environment. I’ll keep the Synology DS1813+ online as an NFS vDP backup datastore.

Feel free to comment and let me know how your experience with the Synology devices using iSCSI MPIO is/was. I’m curious to see if others are experiencing the same results.


UPDATE – June 6th, 2014

The other day, I finally had time to play around and do some testing. I created a new FileIO iSCSI Target, I connected it to my vSphere test environment and configured round robin. Doing some tests on the newly created datastore, the iSCSI connections kept disconnecting. It got to the point where it wasn’t usable.

I scratched that, and tried something else.

I deleted the existing RAID volume and I created a new RAID 5 volume and dedicated it to Block I/O iSCSI target. I connected it to my vSphere test environment and configured round robin MPIO.

At first all was going smoothly, until again, connection drops were occurring. Logging in to the DSM, absolutely no errors were being reported and everything was fine. Yet, I was at a point where all connections were down to the ESXi host.

I shut down the ESXi host, and then shut down and restarted the DS1813+. I waited for it to come back up however it wouldn’t. I let it sit there and waited for 2 hours for the IP to finally be pingable. I tried to connect to the Web interface, however it would only load portions of the page over extended amounts of time (it took 4 hour to load the interface). Once inside, it was EXTREMELY slow. However it was reporting that all was fine, and everything was up, and the disks were fine as well.

I booted the ESXi host and tried to connect to it, however it couldn’t make the connection to the iSCSI targets. Finally the Synology unit became un-responsive.

Since I only had a few test VMs loaded on the Synology device, I decided to just go ahead and do a factory reset on the unit (I noticed new firmware was available as of that day). I downloaded the firmware, and started the factory reset (which again, took forever since the web interface was crawling along).

After restarting the unit, it was not responsive. I waited a couple hours and again, the web interface finally responded but was extremely slow. It took a couple hours to get through the setup page, and a couple more hours for the unit to boot.

Something was wrong, so I restarted the unit yet again, and again, and again.

This time, the alarm light was illuminated on the unit, also one of the drive lights wouldn’t come on. Again, extreme unresponsiveness. I finally got access to the web interface and it was reporting the temperature of one of the drives as critical, but it said it was still functioning and all drives were OK. I shut off the unit, removed the drive, and restarted it again, all of a sudden it was extremely responsive.

I removed the drive, hooked it up to another computer and confirmed that it was failed (which it was).

I replaced the drive with a new one (same model), and did three tests. One with NFS, one with FileIO iSCSI, and one with BlockIO iSCSI. All of a sudden the unit was working fine, and there was absolutely NO iSCSI connections dropping. I tested the iSCSI targets under load for some time, and noticed considerable performance increases with iSCSI, and no connection drops.

Here are some thoughts:
-Two possible things fixed the connection drops, either the drive was acting up all along, or the new version of DSM fixed the iSCSI connection drops.

-While performance has increased with FileIO to around ~120-160MB/sec from ~50MB/sec, I’m still not even close to maxing out the 4 X 1Gb interfaces.

-I also noticed a significant performance increase with NFS, so I’m leaning towards the fact that the drive had been acting up since day one (seeks per second increased by 3 fold after replacing the drive and testing NFS). I/O wait has been significantly reduced

-Why did the Synology unit just freeze up once this drive really started dying? It should have been marked as failed instead of causing the entire Synology unit not to function.

-Why didn’t the drive get marked as failed at all? I regularly performed SMART tests, and checked drive health, there was absolutely no errors. Even when the unit was at a standstill, it still reported the drive as working fine.

Either way, the iSCSI connection drops aren’t occurring anymore, and performance with iSCSI is significantly better. However, I wish I could hit 200MB+/sec.

At this point it is usable for iSCSI using FileIO, however I was disappointed with BlockIO performance (BlockIO should be faster, no idea why it isn’t).

For now, I have an NFS datastore configured (using this for vDP backup), although I will be creating another FileIO iSCSI target and will do some more testing.

Jul 232012

Interesting story:

On the weekend, my Trixbox VoIP PBX (which runs Asterisk) failed. Unfortunately, during the restore process the hard drive also blew up. Temporarily I setup a ML350G5 as a temp VoIP PBX, however today I had the chance to setup an old Acer Aspire Netbook which I had sitting in a box as my new permanent VoIP PBX.

I used a USB CD-Rom to install Trixbox, as for some reason I couldn’t get it to load the kickstart file during a grub boot off a USB key, also couldn’t get it to mount the NFS install export (maybe the kernel didn’t have support for NFS?).


The netbook had decent specs:

Dual-Core 1.5Ghz Process

1GB Ram

Battery (this means I don’t have to put it on my UPS for all my server equipment)


Got it setup and it’s running great! 🙂

Jun 292012

As most of you have read, I received 2 X Raspberry Pi the other day. I’ve been actively hacking and working away on these lovely little devices.

One of the projects I wanted to do, was get Lio-Target (iSCSI Target) running on the Pi. I know that the Pi doesn’t have gigabit networking, but I thought this would still be an interesting proof of concept. Anyways, I got it running, and I have succesfully connected to a USB storage device which was configured as a iSCSI target on my Pi, from my Windows 7 workstation.

This is a brief overview, I will be providing instructions in detail at a later date. Here’s how I did it:

1) Download Fedora 17 for ARM (build for Raspberry Pi).

2) Put latest Firmware and Kernel from Raspberry Pi github repo on to the boot partition. Resize my 16GB card so I have boot, root, and a 2 GB swap.

3) Download snapshot of Raspberry Pi kernel sources. I built the iSCSI Target as modules (I also threw in some other stuff for future projects but it’s not important right now).

3) Install compilers, libraries, etc for kernel build process.

4) Compile kernel

5) Build Raspberry Pi kernel image using Raspberry Pi image tools on github repo, copy to boot.

6) Boot off new kernel

7) Install Target CLI from yum (this was a nice change from compiling on my own), and then build Lio-Utils (this isn’t mandatory, but I like Lio-utils).

8) Configure target, connect, test.

Here’s a copy/paste of proof I have it running!

[root@fedora-arm lio-utils.git]# uname -a
Linux fedora-arm #1 PREEMPT Thu Jun 28 16:40:46 MDT 2012 armv6l armv6l armv6l GNU/Linux
[root@fedora-arm lio-utils.git]# w
09:12:10 up 34 min,  5 users,  load average: 0.99, 1.51, 1.10
USER     TTY      FROM              LOGIN@   IDLE   JCPU   PCPU WHAT
root     pts/0    host.digitall 31Dec69 22:38   2.27s  0.02s tail -f /var/log/messages
root     pts/1    host.digitall 31Dec69  0.00s  1.19s  0.05s w
root     pts/2    host.digitall 08:58   13:24  13.89s 13.04s top
root     pts/3    host.digitall 31Dec69  0.00s  0.00s   ?    –
[root@fedora-arm lio-utils.git]# /etc/init.d/target status
[—————————] TCM/ConfigFS Status [—————————-]
\——> iblock_0
HBA Index: 1 plugin: iblock version: v4.1.0-rc1-ml
\——-> array0
Status: ACTIVATED  Execute/Left/Max Queue Depth: 0/128/128  SectorSize: 512  MaxSectors: 240
iBlock device: sdb  UDEV PATH: /dev/sdb
Major: 8 Minor: 16  CLAIMED: IBLOCK
udev_path: /dev/sdb

[—————————] LIO-Target Status [—————————-]
\——-> tpgt_1  TargetAlias: LIO Target
TPG Network Portals:
\——-> IP-hidden:3260
TPG Logical Units:
\——-> lun_0/30b42bf9f5 -> target/core/iblock_0/array0

Target Engine Core ConfigFS Infrastructure v4.1.0-rc1-ml on Linux/armv6l on
RisingTide Systems Linux-iSCSI Target v4.1.0-rc1
[root@fedora-arm lio-utils.git]# cat /proc/cpuinfo
Processor       : ARMv6-compatible processor rev 7 (v6l)
BogoMIPS        : 697.95
Features        : swp half thumb fastmult vfp edsp java tls
CPU implementer : 0x41
CPU architecture: 7
CPU variant     : 0x0
CPU part        : 0xb76
CPU revision    : 7

Hardware        : BCM2708
Revision        : 0002
Serial          : 00000000c1ad6318
[root@fedora-arm lio-utils.git]#

Jun 092012

So there I was… Had a custom built vSphere cluster, 3 hosts, iSCSI target (setup with Lio-Target), everything running fine, smoothly, perfect… I’m done right? NOPE!

Most of us who care, are concerned about our Disaster Recovery or Backup solution. With virtualization things get a little bit interesting in the fact that either you have a CRAZY large setup, and can use the VMware backup stuff, or you have a smaller environment and want something simple, easy to use, and with a low footprint. Configuring a backup and/or disaster recovery solution for a virtualized environment may be difficult and complicated, however after it’s fully implemented; management, use, and administration is easy, and don’t forget about the abilities and features you get with virtualization.

Reasons why virtualization rocks when it comes to backup and disaster recovery:

-Unlike traditional backups you do not have to install a bare OS to run the backup software to restore over

-You can restore to hardware that is not like nor similar to the original hardware

-Backups are now simple files that can be easily moved, transported, copied, and saved on to normal or non-normal media (you could save an entire system on to a USB key if it was big enough). On a 2TB external USB drive, you could have a backup of over 16+ virtual machines!

-Ease of recovery: Copy the backed up VM files to host/datastore, and simply hit play. Restore complete and you’re up and running!


So with all that in mind, here we go! (Scroll to bottom of post for a quick conclusion).

For my solution, here are some of the requirements I had:

1) Utilize snapshots to take restorable backups while the VMs are running (no downtime).

2) Move the backups to a different location while running (this could be a drive, NFS export, SMB share, etc…).

3) Have the backups stored somewhere easy to access where I can move them to a removable external USB drive to take off-site. This way, I have fast disk-to-disk access to restore backups in the event my storage system goes down or RAID array is lost (downtime would be minimal), or in the event of something more serious like a fire, I would have the USB drive off-site to restore from. Disk-to-disk backups could happen on a daily basis, and disk-to-USB could be done weekly and taken off-site.

4) In the event of a failure, be able to bring USB drive onsite, transfer VMs back and be up and running in no time.


So with this all in mind, I started designing a solution. My existing environment (without backup) composed of:

2 X HP DL360 G5 Servers (running ESXi)

1 X HP ML350 G5 Server (running ESXi)

1 X Super Micro Intel Xeon Server (Running CentOS 6 & Lio-Target backports: providing iSCSI VMFS)

2 X HP MSA20 Storage Units


First, I need a way to create snapshots of my 16+ virtual machines. After, I would need to have the snapshots moved to another location (such as a backup server). There is a free script available called ghettoVCB. ghettoVCB is a “Free alternative for backup up VM’s for ESX(i) 3.5, 4.x+ & 5.x” and is available (along with tons of documentation) at: This script is generally ran on the ESXi host, generates a snapshot and clones it to a seperate datastore configured on that host. It does this for all virtual machines named in the VM list you specify, or for all VMs on the host if a specific switch is passed to

So now that we have the software, we need to have a location setup to back up to. We could either create a new iSCSI target, or we could setup a new Linux server and have a RAID array configured and formatted with ext4 and exported via NFS. This would allow us to have the NFS setup as a datastore on ESXi (so we can backup to the NFS export), and afterwards be able to access the backups natively in Linux to copy/move to a external drive formatted with EXT4.

We configure a new server, running CentOS 6 with enough storage to backup all VMs. We create NFS exports and mount these to all the ESXi hosts. We copy the ghettovcb script to a location on the NFS export so it’s accessible to all hosts easily (without having to update the script on each host individually), and we create lists for each physical host containing the names of the virtual machine it virtualizes. We then edit the file to specify the new destination datastore (the backup datastore) and how we want it to back up.

When executing:

./ -f esxserverlist01

It creates the snapshot for each VM in the list for that host, clones it to the destination datastore (which in my setup is the NFS export on the new backup server), then deletes the snapshot when the backup is complete, finally moving on to the next VM and repeating the process until done. The script needs to be ran on all hosts, and list files for VMs have to be created for each host.

We now have a backup server and have done a disk-to-disk backup of our virtual machines. We can now plug in a large external USB drive to the backup server, and simply copy over the backups to it.


I do everything manually because I like to confirm everything is done and backed up properly, however you can totally create scripts to automate the whole process. After this we have our new backup solution!


Quick recap:

1) Setup a new backup server with enough disk space to back up all VMs. Setup an NFS export and mount it to ESXi hosts.

2) Download and configure the ghettoVCB script. Run the script on each ESXi host to disk-to-disk backup your VMs to your new backup server.

3) Copy the backup files from the backup server to a external USB drive that has enough space. Take off-site.


I have had to restore a couple VMs in the past due to a damaged RAID array, and I did so using a backup from above. It worked great! I will create a post on the restore process sometime in the future (for now feel free to look at the ghettoVCB documentation)!

Jun 092012

Recently, I’ve started to have some issues with the HP MSA20 units attached to my SAN server at my office. These MSA20 units stored all my Virtual Machines inside of a VMFS filesystem which was presented to my vSphere cluster hosts over iSCSI using Lio-Target. In the last while, these logical drive has just been randomly disappearing, causing my 16+ virtual machines to just halt. This always requires me to shut off the physical hosts, shut off the SAN server, shut off the MSA20s, and bring everything all the way back up. This causes huge amounts of downtime, and it just a pain in the butt…

I decided it was time for me to re-do my storage system. Preferably, I would have purchased a couple HP MSA60s and P800 controllers to hook it up to my SAN server, but unfortunately right now it’s not in the budget.

A few years ago, I started using software RAID. In the past I was absolutely scared of it, thought it was complete crap, and would never have touched it, but my opinion drastically changed after playing with it, and regularly using it. While I still recommend businesses to use Hardware based RAID systems, especially for mission critical applications, I felt I could try out software RAID for the above situation since it’s more of a “hobby” setup.


I read that most storage enthusiasts use either the Super Micro AOC-SASLP-MV8, or the LSI SAS 9211-8i. Both are based off different chipsets (both of which are widely used in other well known cards), and both have their own pro’s and con’s.

During my research, I noticed a lot of people who run Windows Home Server were utilizing the AOC Super Micro Card. And while using WHS, most reported no issues whatsoever, however it was a different story when reading posts/blog articles from people using Linux. I don’t know how accurate this was, but apperently a lot of people had issues with this card under heavy load, and some just couldn’t get it running inside of linux.

Then there is the LSA 9211-8i (which is the same as the extremely popular IBM M1015). This bad boy supports basic RAID operations (1, 0, 10), but most people use it with JBOD and simply use Linux MD Software RAID. While there was numerous complaints about users having issues with their systems even detecting their card, other users also reported issues caused by the BIOS of this card (too much memory for the system to boot). When people did get this card working though, I read of mostly NO issues under Linux. Spent a few days confirming what I already had read and finally decided to make the purchase.

Both cards support SAS/SATA, however the LSI card supports 6Gb/sec SAS/SATA. Both also have 2 internal SFF8087 Mini-SAS connectors to hook up a total of 8 drives directly, or more using an SAS expander. The LSI card uses a PCIe (V.2) 8x slot, vs the AOC-SASLP which uses PCIe (V.1) 4x slot.


I went to and ordered the LSI 9211-8i along with 2 breakout cables (Card Part#: LSI00194, Cable Part#: CBL-SFF8087OCF-06M). This would allow me to hook up a total of 8 drives (even though I only plan to use 5). I already have an old computer I already use with an eSATA connector to a Sans Digital SATA Expander for NFS, etc… that I plan on installing the card in to. I also have an old Startech SATABAY5BK enclosure which will hold the drives and connect to the controller. Finished case:

Server with disk enclosures (StarTech SATABAY5BK)










(At this point I have the enclosure installed along with 5 X 1TB Seagate 7200.12 Barracuda drives)

Finally the controller showed up from NCIX:

LSI SAS 9211-8i









I popped this card in the computer (which unfortunately only had PCIe V1), and connected the cables! This is when I ran in to a few issues…

-If no drives were connected, the system would boot and I could succesfully boot to CentOS 6.

-If at all I pressed CTRL+C to get in to the cards interface, the system would freeze during BIOS POST.

-If any drives were connected and detected by the cards BIOS, the system would freeze during BIOS POST.


I went ahead and booted in to CentOS 6. Downloaded the updated firmware and BIOS and flashed the card. The flashing manual was insane, but had to read it all to make sure I didn’t break anything. First I updated both the firmware and BIOS (which went ok), however I couldn’t convert the card from IR firmware to IT firmware due to errors. I google’d this and came up with a bunch of articles, but this one: was the only one that helped and pointed me in the right direction. Essentially just stating you have to use the DOS flasher, erase the card (MAKING SURE NOT TO REBOOT OR YOU’D BRICK IT), and then flashing the IT Firmware. This worked for me, check out his post! Thanks Bryan!

Anyways, after updating the card and converting it to the IT firmware. I still had the BIOS issue. I tried the card in another system, and still had a bunch of issues. I finally removed 1 of 2 video cards and populated the card in a Video Card slot, and I finally could get in to the BIOS. First I enabled staggered spin-up (to make sure I don’t blow the PSU on the computer with a bunch of drives starting up at once), changed some other settings to optimize, and finally disabled the boot BIOS, and changed the option for the adapter to be disabled for boot, and only available to the OS. When removing the card, and putting it in the target computer, this worked. Also noticed that the staggered spin-up started during the Linux kernel startup when initializing the card. Here’s a copy of the kernel log:

mpt2sas version loaded
mpt2sas 0000:06:00.0: PCI INT A -> Link[LNKB] -> GSI 18 (level, low) -> IRQ 18
mpt2sas 0000:06:00.0: setting latency timer to 64
mpt2sas0: 64 BIT PCI BUS DMA ADDRESSING SUPPORTED, total mem (3925416 kB)
mpt2sas 0000:06:00.0: irq 24 for MSI/MSI-X
mpt2sas0: PCI-MSI-X enabled: IRQ 24
mpt2sas0: iomem(0x00000000dfffc000), mapped(0xffffc900110f0000), size(16384)
mpt2sas0: ioport(0x000000000000e000), size(256)
mpt2sas0: sending message unit reset !!
mpt2sas0: message unit reset: SUCCESS
mpt2sas0: Allocated physical memory: size(7441 kB)
mpt2sas0: Current Controller Queue Depth(3305), Max Controller Queue Depth(3432)
mpt2sas0: Scatter Gather Elements per IO(128)
mpt2sas0: LSISAS2008: FWVersion(, ChipRevision(0x03), BiosVersion(
mpt2sas0: Protocol=(Initiator,Target), Capabilities=(TLR,EEDP,Snapshot Buffer,Diag Trace Buffer,Task Set Full,NCQ)
mpt2sas0: sending port enable !!
mpt2sas0: host_add: handle(0x0001), sas_addr(0x5000000080000000), phys(8)
mpt2sas0: port enable: SUCCESS

SUCCESS! Lot’s of SUCCESS! Just the way I like it! Haha, card intialized, had access to drives, etc…


Configured the RAID 5 Array using a 256kb chunk size. I also changed the “stripe_cache_size” to 2048 (the system has 4GB of RAM) to increase the RAID 5 performance.

cd /sys/block/md0/md/

echo 2048 > stripe_cache_size


At this point I simply formatted the drive using EXT4. Configured some folders, NFS exports, and then used Storage vMotion to migrate the Virtual Machines from the iSCSI target, to the new RAID5 array (currently using NFS). The main priority right now was to get the VMs off the MSA20 so I could at least create a backup after they have been moved. Next step, I’ll be re-doing the RAID5 array, configuring the md0 device as a iSCSI target using Lio-Target, and formatting it with VMFS. The performance of this Software RAID5 array is already blowing the MSA20 out of the water!

Here’s some videos of the LEDs on the card in action:


So there you have it! Feel free to post a comment if you have any questions or need any specifics. This setup is rocking away now under high I/O with absolutely no problems whatsoever. I think I may go purchase another 1-2 of these cards!

Jun 062012

So, as most of you know, I have TONS of articles pertaining to getting Lio-Target running on Cent OS. In the beginning, things seemed rather “hit or miss” due to weird errors when either building lio-target or lio-utils…

Turns out, most of the issues I’ve had are related to Python and the current running version. Recently I updated one of my storage boxes using yum, and it completely broke Lio-Target, and Lio-Utils when I had to rebuild them for the new kernel. I was in a panic to mount an old CentOS 6 ISO to get one of the first Python version for CentOS. After downgrading, I was able to build and install both.


Just a heads up for you people getting weird python errors.

Mar 062012

Well, I received a phone call from my father this morning, demanding I to go to his blog and check out a video… (I’ve been helping my dad get this first blog up and running the past couple days).

Check the video out at:, pretty funny 🙂

Anyways, this made me think of some of my old favorite classics. Here’s a few pertaining to Linux:

Linux is Ready


Crime of the century