Jump to content
XPEnology Community

berwhale

Member
  • Posts

    183
  • Joined

  • Last visited

Posts posted by berwhale

  1. On 12/12/2018 at 2:54 PM, NoFate said:

    hi,

     

    ia have several xpenology on my esxi

    the one with PLEX i gave 4 cpu

    can i just give the other ones like 1 cpu only? is it stable?

    or does it always needs to be 2 ? because 3615 is also 2

    1 cpu 2 cores

    or

    2 cpu 1 core?

     

    I wouldn't worry about it too much, ESXi will manage CPU resources across all your VMs and CPU cores.  'over allocating' vCPUs to a little used VM is unlikely to impact your main systems as by definition, it's not doing much.

     

    I allocate 4 vCPU for my main server with Plex and 2 vCPUs for the other DSM instances (downloader, surveillance, test) - this is on a Dell T20 (3.4Ghz quad core Xeon) with 24GB RAM.

  2. Hi Jokerigno, 

     

    It's quite easy to convert from bare metal (Physical) to Virtual (P2V) without loosing any data.  DSM and it's configuration is stored on your data disks, so all you need to do is virtualize the bootloader (Xpenology) - i.e. replace the USB key with a Vmdk attached to a virtual machine.  You then attach your existing disks to the virtual machine and it will retain all of your configuration and data.  If you match the bootloader versions on both physical and virtual machines, then you should avoid any DSM upgrade during the P2V.

     

    When trying this for the 1st time, I was in the fortunate position of having several old PCs and disks lying around, so I was able to create a new bare metal server and P2V that to the T20 before attempting this with my main server (with a fully loaded 4x3TB SHR array).  If you have access to some spare kit, I would encourage you to do the same.

     

    On vSphere/ESXi, there are two ways to present your existing disks to the virtual machine:

     

    1. DirectPath IO - Install a PCI-e SATA adaptor in the T20, connect your existing disks to the adaptor and pass the adaptor through to the new VM.  DSM will see all of the disks natively and will have access to SMART data.  This is the route I went down when P2Ving my server several years ago.  There are some downsides to this route that affect your ability to manage the VM (e.g. All the RAM you configure for the VM is dedicated to the VM and can't be shared with other VMs, you can't take snapshots for the VM for backup purposes)  In practice, these limitations have not been a problem for me.  I think this is the PCI-e card i'm using: https://www.amazon.co.uk/gp/product/B00AZ9T3OU?psc=1&redirect=true&ref_=oh_aui_detailpage_o00_s00 (I seem to recall that the Marvell chipset was important for DSM support at the time I bought it)

     

    2. Raw Device Mapping (RDM) - Disks can be attached to the VM individually as RDMs, this involves creating links to the physical disks on the ESXi command line, then attaching these links to the VM as disks.  This configuration will give you a bit more flexibility from a ESXi perspective, but you do loose visibility of SMART data in DSM.

     

    My own T20 (24GB RAM) is currently running 5 instances of DSM, with a mixture of both approaches above:   e.g. Main file server has PCI-e adapter passed through with several disks attached.  Surveillance server has a single 2TB surveillance drive (WD Purple) attached as RDM.

     

    Here's my own notes on how to pass through a drive via RDM...

     

    Via SSH:

     

    List Disks:

     

    ls -l /vmfs/devices/disks

     

    Identify 2TB Surveillance Disk:

     

    t10.ATA_____WDC_WD20PURX2D64P6ZY0_________________________WD2DWCC4M3XP0D4A

     

    Create RDM mapping:

     

    vmkfstools -z /vmfs/devices/disks/t10.ATA_____WDC_WD20PURX2D64P6ZY0_________________________WD2DWCC4M3XP0D4A "/vmfs/volumes/Samsung_250GB_SSD/RDM/WD_2TB_RDM_1.vmdk"

    Note: 'Samsung_250GB_SSD' is my primary vSphere data store, so this is where I create the RDMs.

     

    Add 'Existing Disk' to VM:

     

    Set RdmFilter.HbalsShared to TRUE in Advanced Configuration

    Assign to SATA adapter 1

    Set Disk Compatibility to 'Virtual'

    Set Disk Mode to 'Independent - persistent'

     

    More here: https://gist.github.com/Hengjie/1520114890bebe8f805d337af4b3a064

     

    I hope this is some help and good luck :)

     

    Regards, 

     

    Berwhale.

     

  3. - Outcome of the update: SUCCESSFUL

    - DSM version prior update: DSM 6.2-23739

    - Loader version and model: Jun's Loader v1.03b - DS3615xs

    - Using custom extra.lzma: NO

    - Installation type: VM - ESXi 6.7 on Dell T20
    - Additional comments: Switch VMNIC to Intel1000e (was Vmxnet3)

  4. 18 hours ago, MooseMan123 said:

    Hey,

     

    I am thinking of following this guide to get my colo storage in better shape. If I wanted to back the VM up, what parts are actually what? I assume the OS is on the actual drives themselves? So if I had a storage failure on my ESXi host, I would just rebuild the bootloader? 

     

    If my VM failed, could I throw the disks in a real Synology and access the data?

     

    Thanks 

     

    Yes and Yes (as long as the disks are passed through or accessed as RDMs)

     

    Several years ago, I migrated a set of disks attached to baremetal Xpenology install to DSM VM hosted on ESXi.  I have the disks attach to a dedicated SATA adapter and this is passed through to the DSM VM,.  DSM has complete control of the disks (they're not visible in ESXi).  I've also tested going back the other way.  You need a motherboard and CPU that supports DirectPathIO to pull this trick off.  I believe it's also possible by passing the disks through as Raw Device Mappings (RDM), I do have a disk passed through to another DSM VM this way, but I haven't tested it's portability.

  5. 8 hours ago, doughnet said:

    Thank you.  synoboot.vmdk is attached to SATA 0:0

     

    I have 2 seperate drives attached to SATA1:0 & SATA 1:1

     

    What type of drives are attached to SATA 1?   What is the Compatibility and Disk Mode set to?  Does the install work if you just attach one of the drives?

  6. 9 minutes ago, doughnet said:

    after i install the PAT file it reboots and says that it has to recover and due to the drives have been moved ... i have 2 drives added on SATA 1 under 1:0 & 1:1

     

    Your 1st drive, Synoboot.vmdk, should be attached to SATA 0.  If you attach it to SATA 1, DSM will try to wipe it during install.

  7. I did a bit more testing...

     

    Copy 1.2GB MKV from main 'production' DSM VM to workstation = 45MB/Sec (Data is hosted on HDDs connected via Marvell SATA adapter which is passed through with DirectPathIO

     

    Copy 1.2GB MKV from 'test' DSM VM to workstation = 80MB/Sec (Data hosted on SSD data store)

     

    Both DSM VMs are at the same version and are hosted on the same vSwitch. So maybe there is an issue with DirectPathIO on the cheap Marvell adapter that I'm using? I wonder if it's worth swapping it out for a cheap SAS HBA off eBay...

  8. At the moment I get 45MB/sec if I copy a large MKV from my DSM VM to my desktop PC. The network is all Gigabit and there are 2 or 3 switches between the host and the workstation (there are two paths as I have load balanced NICs). Transfers were faster when I ran bare metal DSM on a considerably slower PC, IIRC, I could move files at somewhere between 80-100MB/s.

     

    However, I can transfer files to a Windows 10 VM running on the same host at over 100MB/Sec. So I don't think that there's an issue with passthrough on the SATA adapter, but I may have a problem with my NIC or vSwitch config. I tried disabling one of the NICs, but it didn't make any difference.

  9. NET VIEW \\servername failing with Error 53 indicates and error with NetBIOS name resolution, so had a poke about on DSM.

     

    If you go to Control Panel/Network/General/Advanced Settings you'll find the option to enable responses to ARP requests (Address Resolution Protocol). Ticking this box enabled me to successfully enumerate the shares on my NAS with a NET VIEW command.

     

    However, I still can't browse to it as a Computer in Networks (it does show up under other devices, which takes me to the DSM web interface).

  10. Is it possible to take 1 of the 2tb drives out of the array, leave one in the server and format that, then put the other drive in a windows machine, transfer the data back to the xpenology server then add that 2tb drive back into the server? Since the 2tb drives are in raid 1 so mirrored.

     

    That is an option - you could keep one half of the broken mirror as the 'backup'. Bear in mind that once you've accessed either disk independently, they won't be mirrored anymore and you'll have to re-mirror (i.e. re-write all of the data to one of the disks) if you change your mind.

     

    However, if possible, I would still recommend building the new array with all the disks present at the start...

     

    Firstly, if you start with a single disk and then add another, you have to wait for the data to be re-striped across both disks (this can take a day or two). If you add a third, you have to wait again.

     

    Secondly, you can only add drives that are the same size or larger than those already in the array. This means that you will not be able to add the 1TB drive after either of the 2TBs are in the array.

     

    This is why I suggested buying another drive for the backup and starting the new array with all three drives.

×
×
  • Create New...