View RSS Feed

synaesthesia

What a few weeks!

Rate this Entry
3 Comments
by
synaesthesia
, 8th June 2012 at 09:51 PM (2562 Views)
It's been a mixed bag, in the last month or so. Started off with a week off - fantastic BT managed to completely murder a phone line install which should have taken 10 minutes. ISP (Sky) were quite good to be honest, considering they were the man in the middle between me and Openreach they handled it quite well. I hope they kick BT's backside because Openreach really is laughable.
So, back to school and the proper good gubbins has arrived for the first of two new servers - shiney new Intel pedestal server chassis with redundant PSU's and the works. New mainboard (Intel S5520HCR), CPU's (pair of E5620) and 32GB of DDR3 to go in along with the RAID card we had lying around (Adaptec 5805) running 4x2TB Seagate ES drives.
Plan was simple. Each server would run ESXi with 2 guests - a Raid10 system which hosted the datastore for the main SCCM server and users, and RAID1 on the onboard host based RAID which ran one of two DC's which ran AD, DHCP, DNS.
Fail 1. ESXi doth not like HBAs No worries, got up and running on a single drive. Got all 4 virtual machines up and running, SCCM installed and configured to liking, and starting to play with ADMT to migrate from the old system. Dreamy progress. In the meantime, we'd also been shopping for another RAID card to go alongside the Adaptec, ideally one of the intel jobbies that integrates with the mainboard via a special PCIE slot. Supplier let us down unfortunately with no stock in the channel so had to plump for another solution. Get an 8 channel RAID card, and use the 5805 with another older RAID card in the second machine.
Couple of days later, our new card turns up - a shiny new Adaptec 6805. RAID10 set up again, everything transferred, perfect! Right, let's get the RAID1 set up.

Plugged drives in.
Turn on server and fire up the RAID controller thingy.
"Changes found. Accept the changes to alter the settings?"
No dumps me out and leaves me powerless to do anything. So it must be "Yes".

That "Yes" instantly wiped the data on the RAID1 (one of which had my backups on... fool!) and removed the RAID data from the RAID10 array, making it RAID1 instead.
F***!
F***!
F***!
F***!

Right. Time for the weekend. Relax, don't let it panic me. Didn't stop me coming home with a face like Boris Johnson's babysitter though.

Wednesday rolls around. I'd got a HDD with ESXi already installed so stuck that in, hoping it might see some datastores. Wouldn't boot. Bugger - reinstalled ESXi anyway, fired up and yay! after re-creating the RAID10 array it hadn't lost any data! It'd still definitely wiped the single drive, but that's no biggy, I'd only lost the second DC - so one unimportant machine out of 4 was great news.

Lesson learned. I'm not going forward from this point without a backup. Right, where's that spare 2TB drive. Stick it in, let's copy those VMDK's over. Didn't know that regardless of the file system (VMFS5) any copy or move action converts thin provisioned disks to fat. Dammit! That's about 8TB of provisioned space that I just don't have! (despite only currently using about 50GB in total). Lot's of learning, prannying around with VEEAM and my saviour came with learning about VMKFSTools - easy method of copying thin provisioned disks and even converting fat disks to thin.
If anyone's interested in that method, SSH into your ESXi host (obviously enable that first) and do the following:
vmkfstools -i original.vmdk -d thin ../backup/newdisk.vmdk
Absolute lifesaver, although it's always disconcerting when you watch progress creep up very slowly yet disk activity doing absolutely nada. It worked though!

Very happy man. Now I can go in this Monday and start the heavy stuff. GPOs. AD user migration. Mandatory profile setups. Application management. Some proper testing. Can't wait!

We're waiting for the start of Summer before we built the second server as we're reusing one of the RM chassis in the newest of our CC3 servers, which makes sense as it's a decent chassis with redundant PSU and a SATA backplane already. Shifting the second server and DC to it will be a piece of cake! Gotta love ESX
Categories
Uncategorized

Comments

  1. AButters's Avatar
    Ouch. That's why I just buy servers off the shelf from Dell. Certified to work with zero problems just pop the ESXi CD in and 10 mins later your done.

    Also Veeam Backup free edition is great for quickly making ad hoc backups or even ad hoc migrating VMs between servers. I am loving it for £0!!!
    Updated 13th June 2012 at 12:08 PM by AButters
  2. synaesthesia's Avatar
    Great when you've got the money Our budget is next to nothing; at least, it's no difference when we need to get rid of every single OS on site, replace a great chunk of computers and buy in servers & licensing, than it was last year or the year before when just ticking over. The saving of circa £2k per server was more than enough. Besides, going with intel-everything except the RAID cards thanks to availability gives us the same sort of reliability and warranties.

    Some cracking progress though this week despite it being very broken up (day and a half out supporting primaries) with nailing DFS replication, solving a small issue with AD replication not being instant - didn't realise the minimum time for replication between AD sites is 15 minutes, but a small tweak put that back to same-site default of up to 5 seconds. AD structure is *finished*, all security groups *finished*. Base GPOs *finished*. Base Windows 7 image with Chrome and Office 2010 *finished* Some "nice" customisations like school logo for logon pictures done. Libraries cracked and pointing to the right places, many thanks to this cracking post : http://www.edugeek.net/forums/window...ding-wiki.html
    Attempted to sort out the KMS host for Office 2010 licensing but it's got some stupid thing where you can't actually even test it with less than 5 clients trying to activate. Back end is done at least.
    Had to remove and reapply the distribution point on Server 2 due to it using the C: drive as a secondary package/source location so I could force it to the specified drive but no big deal.

    Next week - importing a large batch of users (Year 13's are a great batch to play with as the majority have left now!) to test with. Just need to script shares/DFS structure.
  3. brownbread's Avatar
    Guess it depends on how critical back up is to your business. Worth looking at your business model or infrastructure. There may be other areas where you can streamline and save costs to which you can get a better back up system.

Trackbacks

Total Trackbacks 0
No other blogs currently reference this blog entry
Trackback URL: