1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Other vSphere Noob, questions

Discussion in 'Tech Support' started by Mister_Tad, 17 Dec 2018.

  1. saspro

    saspro IT monkey

    Joined:
    23 Apr 2009
    Posts:
    9,610
    Likes Received:
    401
    Apologies, been offline for a couple of days with a sick little one so just catching up on this thread.
    If you head down the vCentre route, providing you're backing it on fast storage & don't have a crazy amount of hosts, guests, DRS rules etc then it'll run on 4GB RAM & a core or 2 fine.
    Just build it with the specs it wants at the beginning then edit the VM afterwards.
    I generally find that by the time you've enabled failover clustering etc on WS2016/2019 the RAM usage for the host is well above the cost of ESXi + vCentre.

    Or (what I did at home) was grab a cheap 5 year old mac mini, slap in 16GB of RAM and ran it as a management host for the cluster. Let me add a few extra goodies on there as well (oh how I love NFR licenses).

    Lab stuff still runs on another couple of hosts (Ml350's with 192GB RAM, vSAN etc) but I do like having zero downtime on home Prod, keeps the WAF up
     
  2. Mister_Tad

    Mister_Tad Will work for nuts Super Moderator

    Joined:
    27 Dec 2002
    Posts:
    14,070
    Likes Received:
    2,428
    Installed things last night - opted for vSphere in the end.

    ESXi naturally went off without a hitch, but a couple hiccups with vCenter. Installed it on a i5/8GB/240GB thinkpad that will form a permanent addition to the server cupboard. Refused to install initially due to wanting 100MB more memory than the laptop reported (on account of video mem), quick google later added the NO_HARDWARE_CHECKS option. Then wouldn't install because I didn't have a NIC installed - grabbed the USB NIC that I'll be using and plugged that in, and gave it a fixed IP. Then the install rolled back because whilst the NIC was assigned a fixed IP, it wasn't actually connected to anything. So repeated all steps with it in the cupboard on the network and all's well.

    I need to do some networking rejigging before I move forward though. The XG is an edge switch at the moment, so need to logically swap things around with the 16-150, and probably physically too... I'd feel warmer and fuzzier if the core was at the top of the rack directly under the USG. I also need to beef up the cooling in the cupboard so have ordered an extractor fan... getting toasty. I can keep the door shut during normal operations, but the fans in the kit do crank up a bit. Any time I watch a film I need to open the cupboard door though on account of the 13-channel space heater that's also in there.

    I'd love to be able to pick up the bargain-basement servers on ebay for some serious power, but they're all too deep/loud for me. The cupboard is next door to the media room, and whilst it's sound proofed there's only so much that will do. I'm limited in depth by the inset for the centre speaker from the media room as well, hence using a 2-post rack. All that means is grunt needs to come from newer, low-power kit, which tends to mean ££££ :/

    Longer term plan is a pair of short 1U Xeon-D servers with space for 128GB each... 32 threads in a 45w footprint... so much want (sadly also so much pounds)
     
    Last edited: 20 Dec 2018
  3. Mister_Tad

    Mister_Tad Will work for nuts Super Moderator

    Joined:
    27 Dec 2002
    Posts:
    14,070
    Likes Received:
    2,428
    I've not gotten very far with this... spent a few hours yesterday re-racking and re-cabling a bunch of stuff (at least it's super tidy now) with a few of getting things up and running today, and I've hit a stumbling block on either networking or storage...

    Two physical NICs attached currently

    10.x.x.16 - vSwitch0, vmk0, management
    10.x.x.15 - vSwitch1, vmk1, data & VM network
    Create NFS datastore, export to 10.x.x.15 - fail

    My explanation of the setup and the issue highlight what's really the root cause here, in that I'm completely winging it, alas I'd like to carry on plugging away until I get it sussed since IME managing vSphere once it's built up is fairly painless... I've just never built a fresh environment before.

    Any protips?

    EDIT: Meh... seems like management and data interfaces on the same subnet was the problem. Which of course means that the "I'll get to that later" task of separating out a management VLAN is now on the critical path :/
     
    Last edited: 23 Dec 2018
  4. Zoon

    Zoon Hunting Wabbits since the 80s

    Joined:
    12 Mar 2001
    Posts:
    5,869
    Likes Received:
    809
    So you can have management and VM network as the same vswitch, but you have to have a dedicated vswitch for iscsi - two if you're going to do MPIO still.
     
  5. Mister_Tad

    Mister_Tad Will work for nuts Super Moderator

    Joined:
    27 Dec 2002
    Posts:
    14,070
    Likes Received:
    2,428
    They're on physically separate switches, so it made most sense to me to have separate vswitches too.
    Alas, I might can the whole thing and head back for Hyper-V.

    I got started by just being lazy and using the one interface for both management and data and this was fine to create the datastore, got a VM running, relocated scratch to the datastore, set up the management VLAN and moved management over to another NIC... and at some point in there everything went to sh__.

    Following a restart of the host for I can't even remember what, it became completely unresponsive - seemed happy enough from the console, but both the web interface and vCenter just whirred forever when doing anything. It's absolutely because of something I did not knowing better, and I'm sure I could get everything ironed out given some more time, but it just made me consider what if this were 10 months down the line when I don't have free time on tap to troubleshoot, and I make a seemingly innocuous change and everything tanks and I have no idea why.

    At the moment, it's not really matching the faff-free setup I had with hyper-v previously :/
    What I need is a technical support person on-tap... I've turned into my parents :worried:
     
    Last edited: 23 Dec 2018
  6. Mister_Tad

    Mister_Tad Will work for nuts Super Moderator

    Joined:
    27 Dec 2002
    Posts:
    14,070
    Likes Received:
    2,428
    After ingesting a few docs docs whilst loafing over Christmas, I've come back to this and done from what I can tell is the exact same config, and everything is rocking and rolling.

    I can only surmise that the process of moving the management interface, and the dickery that entailed, somehow broke something, as that's the only thing that's different this time around - I wonder whether previously I missed something and the storage ended up going across the router from the management network to the data network?

    Pleased with storage performance as well, as I was wondering how well it would cope...

    [​IMG]

    There may be more there, but I'll get everything set up before doing any performance tweaking, as the SSD cache seems to be sensibly used, and give the best of HDD and SSD simultaneously.
     
    Last edited: 28 Dec 2018
  7. saspro

    saspro IT monkey

    Joined:
    23 Apr 2009
    Posts:
    9,610
    Likes Received:
    401
    All sorted now?
     
  8. Mister_Tad

    Mister_Tad Will work for nuts Super Moderator

    Joined:
    27 Dec 2002
    Posts:
    14,070
    Likes Received:
    2,428
    For now. Until I migrate the data network from a standard to distributed vswitch and break things all over again.
     
  9. saspro

    saspro IT monkey

    Joined:
    23 Apr 2009
    Posts:
    9,610
    Likes Received:
    401
    That's not too bad to do. Tip - Leave the management network on a standard vswitch, you can add another management network later on the Dswitch but leave one on the standard one to begin with
     
  10. Mister_Tad

    Mister_Tad Will work for nuts Super Moderator

    Joined:
    27 Dec 2002
    Posts:
    14,070
    Likes Received:
    2,428
    I don't think there will be any reason for me to be moving the management across so will leave it well alone, as that feels like what hosed me last time.

    I think I've got it all set up to just migrate now, with perhaps a momentary blip to configure LACP on the switch. I need to shut down the NAS to move it down 1U to make way for a US-48 next week, so will do it all in the same change window.
     
  11. Zoon

    Zoon Hunting Wabbits since the 80s

    Joined:
    12 Mar 2001
    Posts:
    5,869
    Likes Received:
    809
    I hope you have your full ITIL-based Change Management Record logged and approved with your service owners.
     
  12. Mister_Tad

    Mister_Tad Will work for nuts Super Moderator

    Joined:
    27 Dec 2002
    Posts:
    14,070
    Likes Received:
    2,428
    I'd like to say I arranged the appropriate outage windows as part of the investment review board, but I was the only one at that meeting.
     
  13. Zoon

    Zoon Hunting Wabbits since the 80s

    Joined:
    12 Mar 2001
    Posts:
    5,869
    Likes Received:
    809
    You've already broken ITIL, you need independent change review and approval! You can't submit and approve your own change!
     
  14. Mister_Tad

    Mister_Tad Will work for nuts Super Moderator

    Joined:
    27 Dec 2002
    Posts:
    14,070
    Likes Received:
    2,428
    ITIL does say that one person can hold several formal roles... I'm just being efficient and limiting ticket-passing.
    I'll make sure I'm wearing all of my ITIL pins before I do any work just to be sure :lol:
     

Share This Page