Results 1 to 4 of 4

Thread: VM and storage server question(s)

  1. #1
    Super MURCer TransformX's Avatar
    Join Date
    Sep 2003
    Location
    Behind you
    Posts
    4,985

    Default VM and storage server question(s)

    hi,

    We're considering upgrading our VM and storage infrastructure here at work and we need stuff to be on line 24/7 bla bla bla.
    the idea we'd like to implement, if possible is the following:
    • Two identical servers for combined VM (~8TB) and Storage (~15TB file server as shared mounts for different teams).
    • We split all the virtual machines and storage to two portions, call them left and right.
    • Both servers hold everything (left and right combined)
    • One server serves left, the other server serves right and they sync their active half to the other server.
    • In case one of the servers fail, the remaining one takes over both sides.


    Now for the questions:
    1. Is this ideal? Is there a better, safer, cheaper implementation that would allow for high availability?
    2. What is the best/recommended software platform?


    Cheers!
    "For every action, there is an equal and opposite criticism."

  2. #2
    Super MURCer MultimediaMan's Avatar
    Join Date
    Sep 1999
    Location
    Bentonville, Arkansas, USA
    Posts
    1,874

    Default

    Gluster or Ceph on Linux can present SMB/NFS and Raw iSCSI LUNs (as files) to Virtual hosts as well as serve files. The ideal number of servers is four (or more)... but these servers can be smaller. There are solutions for VMware by VMware (vSAN), HP (Stor-Virtual aka "Lefthand"), Nutanix (AHV). vSAN and Nutanix requires 4 or more storage nodes for proper striping and fault domains. Gluster and Ceph are likewise set at four for Striping/ Fault Domain/ Tiebreaking.

    If you roll your own, you run into problems with Synchronization (rsync & robocopy) only take you so far), and their are problems with DNS/ Name Resolution/ IP addresses when one side dies, and we haven't even tackled such trivialities such as storage performance, RPO, and RTO...

    More information is needed.
    Last edited by MultimediaMan; 3rd November 2019 at 09:29. Reason: More information...
    Hey, Donny! We got us a German who wants to die for his country... Oblige him. - Lt. Aldo Raine

  3. #3
    Super MURCer TransformX's Avatar
    Join Date
    Sep 2003
    Location
    Behind you
    Posts
    4,985

    Default

    Hey MultimediaMan!

    Thanks for the quick reply! I'll study the solutions you've mentioned. Meanwhile I'll try to provide more information:

    File storage wise, we have different teams (~200 people maximum in total) mounting (SMB) 'drives' on their windows machines and collaborating.
    We're talking about Microsoft office files of all kinds, media files etc. Not databases or high IOPS.
    No network drives, no work.

    Please feel free to ask more.
    "For every action, there is an equal and opposite criticism."

  4. #4
    Super MURCer UtwigMU's Avatar
    Join Date
    Jul 2002
    Location
    Sin City
    Posts
    5,087

    Default

    I finally moved servers from crappy inaccessible closet to a nice server rack in a dedicated server room and I'm pondering similar things. I have two ESXes which host VMs, newer full, older half empty, hosts less important services and my Linux test servers.

    For the firewall and VLANs I brought in external consulting company and they are really big on Proxmox (linux based VM host using ZFS on linux which can snapshot and replicate). I still need to tidy up a few things but I'll be definitely looking at proxmox since now I don't have good high availability and quick disaster recovery. I backup windows VMs to ISCSI on the NAS. I'll be ordering new host sometimes spring.

    I'm also thinking of getting a dedicated file server and leave VMs (300-1TB) only for the apps.

    And my users are also too inept to discover network neighbourhood and access shares, while boss is so creative that we have more shares than users. So I too map mostly used shares with script to respective users and give people shortcuts for rarely used shares.

    For failover you need storage and VM host without storage (server mobo with CPUs) rarely fails, drives are more likely to fail, so you might want redundant storage. With storage except for running storage on VM host or running a virtualized ZFS NAS with bare metal exposed drives it gets into mid 5 digits pretty quickly (20-50k).
    Last edited by UtwigMU; 3rd November 2019 at 15:43.

Thread Information

Users Browsing this Thread

There are currently 1 users browsing this thread. (0 members and 1 guests)

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •