Here I share a couple of issues I have ran into during a conversion of a physical server to VMWare ESXI server that I hope could help someone else.
The good news is, I figured out a lot of stuff that I needed to do when I make the next attempt. The main problem I am having is the customer’s downtime. They couldn’t be down an extra day (like I told them to be prepared for), so next time I am going to make sure I have at the very least 3 days to work with (to iron out any issues).
More details in the video and in the section below the video:
Some useful tips when working with datastores and swap in ESXi
Here are a couple things I found along the way:
1) When creating the datasets, if it’s over 256GB of space, you need to adjust the block size (for example, set the block size to 1 MB if its under 256GB, between 256 GB and 512GB, set it to 2 GB, and so on). If you don’t it will show you that the dataset is 450 GB (or whatever size it really is), but only allow you to utilize up to 256GB. I didn’t realize this until I was pretty far in the process.
A block size of 1MB = 256GB maximum file size
A block size of 2MB = 512GB maximum file size
A block size of 4MB = 1024GB maximum file size
A block size of 8MB = 2048GB maximum file size
2) ESXI needs space on one of your datasets for swap. It wants the same GB size as your hosts’ RAM. It also wants it created on the same dataset as your ESXI installation. This was a problem for me, because when the customer’s hardware vendor sold him the server, they set it up with about the same amount of HDD space as the old one. The problem with that, is that the old one was running a bit low on space on it’s OS partition. When I ran the conversion on the old server, I told it to create a 145 GB dataset (that way the os had a little bit of room for future expansion). That was the size of my RAID1 dataset on the new server. Now, since ESXI needs at least 12GB of room on the dataset for it’s SWAP, I was in trouble, because I dedicated the entire 145GB to the VM. This opened up a whole new can of time wasting… Let me explain:
You can configure ESXI’s SWAP to be located somewhere else other than the default location, so I moved it to my 2nd dataset (remember, at this point I still had it configured to use 1MB size, so it was only really able to use up to 256GB – in which I wasn’t aware of yet). Now I could boot the new VM without any errors. After spending a couple hours restarting the vm, configuring IP, etc…. (also, the server was very slow to respond to anything when booted, thats why it took a couple hours), I then tried to configure the 2nd dataset on the new VM. This would be for my 2 data partitions. Problem here was, I couldn’t see all my 450GB because of the 1MB block size. So…. I removed the 2nd dataset, re-created the 2nd dataset with the 2MB block size. That took care of the not seeing the full 450GB of space, but… created a whole new problem. Since I told ESXI to use the 2nd datastore as it’s partition, now the new VM would not boot. If I move the SWAP config to the 1st dataset, that would work, but remember, I didn’t have any room on that dataset to work with. Also, because the 2nd dataset was gone (the original one), trying to configure the SWAP location would error out, no matter what… Until I found an “advanced” section, where I can manually configure it.
(SIDE NOTE): The reason 1 1/2 days wasn’t enough time, was due to how long the copying of the 450GB of data drives to a slow USB external HDD took. I have to come up with a better solution next time, maybe copy them to a network location.
In the end, I ended up moving the VM from dataset1 to dataset2 so that I could then configure the SWAP to use the default location (on the same dataset as it’s VM). That process said it was going to take 800 minutes to complete. It was 10:00 PM by this time, and the customer didn’t want to be down another day. Time to make some tough decisions… In the morning I went back out there, the move process was done, but I would have had to move it back over again, which would take another good 800 minutes or so. Also, I would have had to iron out any other issues that would have come up.
Now that I have learned a few things… I will have to make another attempt to get this thing migrated over. During my testing of the new VM, I did notice things were working as normal as far as DNS, DHCP, AD, etc… so thats some good news.
Also, I need to talk to the customer, and suggest we purchase 2 larger HDDs for the RAID 1 (the OS RAID), for future expansion. Then I could take the 2 drives that are there now, and move them to the RAID5 (we need more space for the data drives too). This would allow for us to have a large enough RAID1 to put the ESXI SWAP partition on, and have room in case we increase the RAM later on.
I hope the information above is able to help someone else out in a similar situation…. lessons learned for me.