Posts Tagged ‘xserve’

Next, on YOTG.

November 15th, 2011

It has finally happened. My lab has finally gotten the funds to upgrade our aging G5 servers with nice, new, shiny Mac Mini’s and a Promise Pegasus RAID.

Currently, the lab I run has 10 Mac Pro desktop’s all running into a 6 year old G5 Xserve and Apple RAID. The RAID uses 14 IDE drives that are basically maxed out. We have about 4TB of storage on 14 drives..this is very sad. Our G5 servers are not upgradable anymore, and we have limitations on the types of things we can serve on them. They have lasted us this long, but it’s time to finally phase them out.

In the next 2-3 weeks I’ll be replacing our two G5 servers and our RAID with two Mac Mini servers and the Promise Thunderbolt RAID. The servers will give us huge boost in performance and the RAID will bump us to 12TB of storage. This will not be an easy task as our current systems all run 10.5 and the new servers run 10.7. I will also have to migrate all of the user accounts and data to the new system without losing anything. Instead of removing our old servers I will use them only as basic servers; MySQL, Apache, Xgrid controllers, etc. I’m also going to use them as tertiary backups for our user accounts and servers (backing up the new machines and user accounts to the RAID once per week).

Over the next few posts I will attempt to document the migration. I’ll start with initial setup then go to migrating data/accounts then end with the final phase out process. I hope that the next few entries may help people who get into a similar situation as well as keep a record for myself on any problems I might face.

No comments »

Posted in Hardware, Work

Tags: apple Hardware new purchase raid xserve

Imaging a Lab with DeployStudio

September 2nd, 2011

Imaging is a great thing..it really is. When you have more than 2 computers, imaging becomes your best friend…and if you buy new machines or one of your older machines dies or gets messed up, it saves so much time. I have a complete backup ready to deploy at all times for both Mac and Windows.

I image my lab once a year. This ensures that I have the latest updates for every machine for all programs, but it also cleans out the old stuff from the previous year that builds up over time. Apple makes imaging very simple by installing a NetBoot feature on all of their machines and including a NetBoot server installed with MacOS X Server. In the past I used Bombich NetRestore, a free AppleScript based program that helped make NetBoot image sets and helped with deploying them. Mike Bombich stopped making NetRestore and suggested everyone to try DeployStudio for imaging..so I did. I must say that DeployStudio is an amazing program especially for a free program. It’s also very simple to get running and fairly robust. In this post I’m going to go over image creation, setup, and deployment with DeployStudio (DS) and also go over some issues I encountered and how I fixed them.

Creating the NetBoot Set

The first step to any Mac NetBoot is the NetBoot set. What the set is is a basic image file that includes all the tools your computer will need to read the image, copy the image, and even run checks on your computer even if you’re not imaging. It’s a very basic MacOS install that resides on the server. DS creates these images for both PPC and Intel machines in the same set, so any Mac can boot from the same set. After installing DS on your server you can open the DS control panel and begin setting up your system AND create your NetBoot set. I will not be going over server setup in this post, I may save that for a later time.

The DeployStudio Control Panel

When you open the control panel you should launch the assistant (you can also find it in /Applications/Utilities). When the assistant opens you select “Create a DeployStudio NetBoot set and continue. If you’re running the assistant on a computer other than a server you will see this:

DeployStudio DHCP Setup

If you plan on using a server to do the deploying, you can skip this, if not, you’ll have to setup a DHCP server. This depends on your setup, for my case I can skip this. The next step allows you to name your set; set the name and unique identifier to whatever you wish, (unless you have multiple NetBoot sets). When you click continue you will tell the set where the computer should log in and look for the images and workflows.

My settings...

more settings...

The settings above are MY settings, yours will be different. The login and password for mine are supplied by the LDAP server. The final step is the actual save location and creation of the image. Pretty self explanatory. It takes about 5-10 minutes.

Completed NetBoot .nbi file

After image creating is successful you’ll have a nice .nbi file in your save location. This file is basically an image file that contains the bootable images for PPC and Intel as well as the basic MacOS system with some basic utilities like Disk Utility, Terminal and Startup Disk. It’s roughly 2.5 GB and it should be placed on your server in the NetBootSP0 folder (It’s located in [Volume]/Library/NetBoot/). Inside the NetBootSP0 folder will be other folders which DS created during install, these contain various other things for DS and also house your images. I will go over image creation next. This is where we will be able to test to see if your NetBoot Server and set are both working.

Creating Images with DeployStudio

Creating the images is an extremely simple task once you know what settings you need. I will explain the setup with my current settings but attempt to go over most of the other ones.

To start the process, boot your mac and hold the ‘N’ key down during power on, this will perform a network boot (REMEMBER: Your computers must all be on the same subnet, this is the only way to do this without messing with a lot of things!) If your computer boots to the DS screen you will see the DS Runtime Window.

This window shows all of your available jobs in DS. There are a few default jobs that come with DS, we’ll make our own later for deploying. For now we’re gong to select “Create a master from a volume.” Click the Play button at the top and you will come to the heart of the Image creation.

My Image Settings from a PowerPC computer

This window is probably the hardest window we’ve seen so far. First thing is to choose which drive you will make an image of from the dropdown menu. I’ll start with my MacOS partition. After selecting the correct partition I name the image something like 2011_09_02_Intel_lab and leave other settings alone. The keywords are not very important unless you have a lot of images. I usually select Compressed for the type because it saves space and it gives a much faster restoration. Access group is what you would have set in your initial DS setup that I did not cover.

Format is what kind of image you are making. Since I’m doing a MacOS install the Format will be HFS+. I normally select “Auto Detect” but if you want to have HFS+ Journaled, Case-sensitive or both you may want to change it because it will always auto-detect HFS+ without journalising.

Once my settings are correct I click the Play button at the top and the image making process begins. This will take a lot of time depending on the size of the image being created, a 100+GB image will take roughly 2 hours (sometimes more, sometimes less, depends on the machine and network) and it will then compress the image (my images get compressed to about 75GB from 128GB…compression rocks!).

Masters in the NetBootSP0 Folder

After image creation you will see the .dmg file in your NetBootSP0/Masters/HFS folder. (Note: I just found out that new versions of DeployStudio won’t show your images in DS Admin unless you have .hfs in filename before the .dmg, it will automatically add them during image creation, but if you have old images, just add the .hfs right before the .dmg extension).

You can use this same process to create NTFS, FAT, and EXT4 images. Follow the same steps but make sure you leave the Format as “Auto-Detect.” After creating a NTFS image it might take some time to show up in DS admin, this is because some server-side tasks may need to be done, it will show up when that is complete. NTFS imaging requires a little more setup in DS admin beforehand…again, I will not be covering that in this post.

Making Workflows to Deploy Images

DeployStudio comes with an administration program where you can manage images, workflows, packages, scripts, and see progress of NetBooted computers. You can also set up all of your computers in it before hand (names, network settings, licenses, etc) and set up automation for all of your systems so if you want a computer to automatically format and re-image when you NetBoot it, you can do that (please don’t think that’s a great idea…just saying). To start setting up workflows you’ll need to open DS Admin, it’s located in /Applications/Utilities. Enter your server credentials and you’re presented with the DS server information.

The window that opens first is the current (or previous) activities. In this window you can watch and control the computers that are currently working in DS. ou can also see what jobs they were doing, and how far along they are. This screen is very helpful when you have DS running on many machines.

I am going to explain how to setup a dual-boot Mac workflow. The default jobs are very helpful at getting you started, I’m going to start from scratch. To create and edit workflows we’re going to select “Workflows” from the left sidebar and begin setting up our job. Click the “+” button at the bottom and you will be presented with a new blank job. Then click on the little “+” button next to “Drop tasks here.”

Creating a new workflow

The first thing to do is to drop the “Partition a disk” task from the left side to the drop space. Then you should select “Mac OS X + Windows” from the Apply layout template dropdown menu. Resize the partitions to suit your needs, make sure your images will be able to fit on the partitions you make for your drive. I normally do 75% Mac OS/25% Windows, I also normally Automate this process, your mileage my vary.

The next step is to drag the “Restore a disk image” job from the left and drop it after the partitioning job. Your MacOS image should ALWAYS be first of else it will not work. Select “Enter value…” from the Target volume section, then select the “MacOSX” option from the menu. Set your Image to HFS and select the appropriate image from the menu (the one you created earlier). Now, for the options you can read from the image below how to set those. If you’re imaging Mac OS 10.7 Lion you should check “Restore system recovery partitions” but I don’t need this.

My HFS Settings

You may also notice Multicast settings, you can set this up if you’re brave, I don’t need it so it’s ignored. Your HFS partition is complete, now on to Windows.

Drag the “Restore a disk image” job from the left and drop it after the first restoring task. Select “Enter value…” from the Target volume section, then select the “WINDOWS” option from the menu. Set your Image to NTFS and select an appropriate image from the menu again. Settings for Windows is relatively the same as HFS with some exceptions; you should check “Expand restored NTFS partition” and uncheck “Set as default startup volume” unless you want to have Windows as your default. You’ll also notice that all of these tasks are automated, this is so you can boot the computer, select the job, and walk away without intervention.

DS NTFS Settings

Now, you can add more jobs to the workflow such as AD binding, or software updates, but this setup is the basic setup for a dual-boot deploy. Now just rename the job by clicking the name in the top with the other jobs and rename it, you can also add a short description of the job. Your workflow is now complete! Now it’s on to the easiest task…deployment!

Deployment

I say this is the easiest part because it really is. If you have everything set up properly, you should have no issues.

To deploy the image to the computers, boot the machines again pressing the ‘N’ key, when the machine boots to DS you can select the newly created Workflow and press the play button. If you automated everything, that’s it..it will partition your drive and load the images to those partitions. After the job is complete your computers will either tell you it was successful (or failed…more on that below) or they will reboot. If the task was successful, GREAT! Reboot the machines, they will run the final scripts in MacOS then reboot again…MacOS is done. You only have one more thing to do and that’s configure Windows. I won’t go into this because it’s going to be different for everyone, but you will have to activate windows and any other programs that require it because Windows will not keep the activation after imaging.

Issues?

Now, not everyone will be so luck to have a successful run…if you run into any issues visit the DS forums, they are very helpful and pretty speedy. I had one issue that just drove me nuts. When I ran my deployment script the MacOS partition would go fine but once Windows hit it would fail…everytime. DeployStudio keeps logs for every computer on the server, so I took a look and noticed the following errors:

[Thu Sep 1 14:41:15] dyld: unknown required load command 0×80000022
[Thu Sep 1 14:41:16] -> invalid starting block value () defined in MBR for partition /dev/disk0s3.
[Thu Sep 1 14:41:16] Check your partition map. You need to define at least one DOS/FAT partition in order to get the MBR automatically in sync with GPT.
[Thu Sep 1 14:41:20] -> Restore action completed.
[Thu Sep 1 14:41:20] Restoration failure (elapsed time: 0.24 minutes)

I posted in the DS forums (topic link) and in a matter of hours the admin of the forums posted a solution:

Sounds like the custom fdisk command fails on 10.7 DSS netboot sets. You may try to remove the one located in your netboot folder at /Applications/Utilities/DeployStudio\ Admin.app/Contents/Frameworks/DSCore.framework/Resources/Tools/fdisk.

So I tried this and BOOM, successful. It’s great when a developer helps with products so quickly…and I’ve only usually seen this with free or open source projects. So if you’re having issues, the forums are key.

I hope this post helps people out with Mac imaging and deployment. If you have any other questions or issues feel free to ask in the comments. If this post helped you or think it will help others please feel free to repost and share away!

1 comment »

Posted in Guides, Software

Tags: deploystudio free macosx servers software windows xserve

Upgrading from 10.4 Server to 10.5 Server

June 1st, 2009

As I’ve mentioned before, I manage 2 Apple Xservers and a lab full of Macs. When classes ended for the summer I decided to upgrade our Tiger server to Leopard. This was a task that I knew was going to be a lot of work especially because I was consolidating our old Dell web/mail/DNS into the Leopard machine. These computers were set up well before I came into the picture, and I was still learning new things about them as time went on. For example: I didn’t know that the old system (Wanda) ran another DNS on top of the university DNS. Confusing.

So, I decided to do this some time ago when I realized that every time the server would restart passwords would be locked out. This was an issue with the Kerberos setup in Tiger which was half broken by the time I got here. Long story short, I wanted to start over, so I did.

Upgrading from one system to another on a server s never an easy task, and Leopard was definitely no exception at all. I’ve upgraded Linux systems much easier than this. I decided to make the image in the lab one of the PowerPC G5s we have (the server is a G5). Here’s a fairly quick rundown of how I got everything to run like I wanted to.

Install 10.5 Server fresh selecting “Advanced Setup” from the choices offered. We’re going to be hosting a ton of services on here and I want to be able to configure every single one of them myself.
Install ALL updates available BEFORE SETTING ANYTHING UP! This is something I always do. it’s so much easier to start setting up the server after it’s completely updated. I don’t want to set up then have something break because the new update fucked it up and believe me, it happens. During this time I also updated XCode to 3.1.x so I can build anything with MacPorts when needed.
Next I needed to re-install or re-compile PHP5 so I had GD and MCRYPT. Instead of recompiling it I decided to install the Entropy PHP build which comes with all that jazz installed plus a few more things. All I have to do is disable the apple libphp5 plugin that comes wth Apache2. Done.
Next I installed the Autodesk Maya FlexLM license server. This was a bit annoying since there technically is no Leopard build from Autodesk. I decided to download the License server from Autodesk and go into the installer package and snoop around a bit. I found a conf file that I could edit to allow it to avoid checking for 10.5. That allowed me to install part of it, but the license server wouldn’t go. I finally just went into the package and took the launch daemon out and the library and just copied the files to their respective places in the system. Rebooted and it worked fine.
I decided to setup phpmyadmin as well as Webmin for administration. Webmin is great, I’m very used to it and it’s very lightweight. Phpmyadmin is obvious. I copied the folders to /etc and redirected the links in Apache using server admin and ran the setup scripts. I also installed the OS X Web-based open directory password changer while I was at it. All worked fine.
Next I did an rsync from the old webserver for the website. Copied everything over, dumped and restored the MySQL databases and users and voila! Web is back.
I rsync’ed the Mailman mailing lists from the old server. This was another issue. The data, archive, and list confs were all compatible; however, the symlinks on the copied lists were not working because they were still pointing to the old server model. I couldn’t really change the links without mailman complaining so I did some workin around this. First, I copied the archives from the old server to the new, this worked fine. I then exported the lists using server admin export AFTER I copied the lists from the old server. This gave me all mailing lists but nothing else, which is what I needed. I removed all the old lists and confs from the system but left the archives in place. I re-imported the lists that I exported from server admin and BING! All mailing lists recreated the links and all lists were back to their working state. Archives came over, user lists came over, and no complaining mailman.
After that was set I copied all of the DNS settings from the old server and put them in the new DNS. I also gave an updated list to Poly’s IS and disabled the now deprecated IDMI nameserver. Everything was now BXMC. This was about the time that I finally turned off Wanda, the old webserver, for good. Wanda will be reconfigured for Hans.
I enabled everything after I turned off Wanda; DNS, Firewall, DHCP, Web, Mail, etc. DHCP and Firewall were imported over from the Tiger install, so that was nothing.
I attempted to restore the open directory database from the Tiger backup. It wasn’t working, at all. I was getting frustrated because now I was thinking I had to put every user in by hand. I finally figured out that I was attempting to restore the LDAP database using the new kerberos realm and new LDAP searchbase. This won’t work because the backup is using the old one, they just won’t talk. I tried to restore using our old searchbase and realm and it worked! I did attempt to change the realm and searchbase with no success, so for now I’m leaving it at IDMI. No one will see this except myself and the systems so I could not care.
With OD working again I tested all logging into the wiki, website, lab machines, afp, etc. All worked after a password update (which all users will need to do unfrotunately).
The final bit was to get the NetRestore working on the server. I wrote about using NetRestore by Mike Bombich here, and how it reached EOL. I stated that I would continue to use it until it stopped working….well, it stopped. I was sad to see this actually, but our mac’s wouldn’t boot from the system anymore. They all failed for some unexplained reason. I decided to give the one Mike Bombich recommended a shot, DeployStudio. Installing DS was easy, just a simple pkg file. Setup was even fairly easy. I ran the server setup on the server and that’s really it. I used carbon copy cloner on a labmac to create an image (even though DS allows you to create one in the NetBoot screen, it didn’t work) and the NetRestore was back working. I actually would like the wrote more about DS in the future. I am very surprised with how well it works and the features it has built in. But that’ll be another story.

So this entire process took me 4 fresh installs on a lab machine before I got the image the way I wanted it (about 4-5 days). It took 45 minutes to copy the image from the external hard drive to the server, and took about 3 days of tweaks to get everything running smoothly. The entire experience was pretty painless, but it kept me busy for the entire time. I will be re-imaging the secondary server again (it was upgraded to Leopard before the main server last summer) so that I will now have everything fresh and clean for the next semester. Here’s a nice shot of server admin with everything running nicely:-)

If anyone reads this and needs help with a server setup and/or upgrade from Tiger to Leopard please feel free to comment or ask me any questions about it because I’ve seen every issue now with the server and I now know how to get around a lot of things with it.