Solved

Some questions about AFS

  • 16 November 2016
  • 12 replies
  • 7840 views

Userlevel 3
Badge +17
Hello Nutanix support,

Currently, I have some questions related to the AFS operation and architecture, could you please give me an answer as well as make it clearer in explanation?! Thanks in advance

1. When a client trying to access its data in the AFS share, that client will be connected to the File server cluster represented by an FSVM master or connected to one of the FSVM that responsible for the client data VG?. How can I verify that which FSVM the client is connecting to?

2. The data of client will be distributed across VGs created by AFS or just reside in only one VG in one node?
Could I verify the data of client in that/those VGs?

3. Whether the client data integrity behavior is also the same with any VM data? I mean that data will be copied and back up to other VGs across the FS cluster?

4. The connection and request between client and client data just like the below diagram?

client request CVM FSVM VG?!

5. How can I prove to the customer that Nutanix leverage its own techs and architecture that has lots of benefits and advantage in performance?

6. How can I manage the FS cluster by command and/or dive deeper into FS cluster configuration such as which node containing the FS VGs, in case of one FSVM down, how can I verify that other FSVMs can take over its IP as well as client request...?
icon

Best answer by dlink7 17 November 2016, 17:45

The backing disks were the user is sitting is on 1 of the 40 TB volume groups.

You can't delete the volume groups. It's a NO TOUCH ZONE :-)

FSVM is no different than any VM using Acropolis Block Services but it was automated failover - just read the nutanixbible on ABS - http://nutanixbible.com/
you can starting at the below section
Block Services
The Acropolis Block Services (ABS) feature exposes back-end DSF storage to external consumers (guest OS, physical hosts, containers, etc.) via iSCSI.
View original

12 replies

Userlevel 4
Badge +21
Hey Hey All my answer are in the context of 5.0

1. When a client trying to access its data in the AFS share, that client will be connected to the File server cluster represented by an FSVM master or connected to one of the FSVM that responsible for the client data VG?. How can I verify that which FSVM the client is connecting to?

You will connected to where the data is located. The AFS master has nothing to do with it. A general share is always located on the same FSVM. The home share has all of the top level directories(TLD) shared amoung all of the FSVM. You can check out which folder is located where by checking out the DFS tab on the TLD.

2. The data of client will be distributed across VGs created by AFS or just reside in only one VG in one node?


Could I verify the data of client in that/those VGs?

General Share has 1 one volume group assigned to it with 4 disks for data, 2 disks for metadata. Those vdisks can be spread amoung all of the CVMS. ADRS will balance the load from the back end.

The home share has the same expect it is made up of the 15 volume groups.

3. Whether the client data integrity behavior is also the same with any VM data? I mean that data will be copied and back up to other VGs across the FS cluster?

Not sure i get this question. Nutanix DSF will take care of inegrity

4. The connection and request between client and client data just like the below diagram?

client request CVM FSVM VG?!

client request FSVM CVM


5. How can I prove to the customer that Nutanix leverage its own techs and architecture that has lots of benefits and advantage in performance?

Test you can run ---
  • Metadata intensive workloads[list]
  • Home Directories with FSCT
  • User Profiles with LoginVSI
  • Spec SFS2014 SWBUILD
  • Throughput/IOPS intensive workloads
    • FIO ZFS
    • FIO Windows
    • Spec SFS2014 VDA
    [/list]

    6. How can I manage the FS cluster by command and/or dive deeper into FS cluster configuration such as which node containing the FS VGs, in case of one FSVM down, how can I verify that other FSVMs can take over its IP as well as client request...?

    Probably not what your locking for but we intenionally try to had this detail from users because you should have to manage it.
    ncli> file-serverCommand Format:ncli> [=] ...
    Entity:file-server | fs
    Actions and their corresponding Parameters:delete | remove | rmuuid (Req):getuuid (Req):projection (Opt):get-builtin-groupsuuid (Req):projection (Opt):join-domainuuid (Req):windows-ad-domain-name (Req):windows-ad-username (Req):windows-ad-password (Req):organizational-unit (Opt):overwrite (Opt):preferred-domain-controller (Opt):leave-domainuuid (Req):windows-ad-password (Req):windows-ad-username (Opt):list | lscount (Opt):filter-criteria (Opt):search-string (Opt):projection (Opt):protectuuid (Req):activate | activateuuid (Req):dns-server-ip-address-list (Req):ntp-servers (Req):windows-ad-domain-name (Req):windows-ad-username (Req):internal-virtual-network (Req):external-virtual-network (Req):windows-ad-password (Req):internal-virtual-network-gateway (Opt):internal-virtual-network-mask (Opt):internal-virtual-network-ips (Opt):external-virtual-network-gateway (Opt):external-virtual-network-mask (Opt):external-virtual-network-ips (Opt):overwrite (Opt):organizational-unit (Opt):preferred-domain-controller (Opt):create | addname (Req):dns-server-ip-address-list (Req):ntp-servers (Req):windows-ad-domain-name (Req):size-gib (Req):windows-ad-username (Req):internal-virtual-network (Req):external-virtual-network (Req):internal-virtual-network-gateway (Opt):internal-virtual-network-mask (Opt):internal-virtual-network-ips (Opt):external-virtual-network-gateway (Opt):external-virtual-network-mask (Opt):external-virtual-network-ips (Opt):nvm-count (Opt):memory (Opt):cpu-count (Opt):windows-ad-password (Opt):organizational-unit (Opt):pd-name (Opt):preferred-domain-controller (Opt):edit | updateuuid (Req):total-nvm-count (Opt):internal-virtual-network-ips (Opt):external-virtual-network-ips (Opt):dns-server-ip-address-list (Opt):ntp-servers (Opt):size-gib (Opt):memory (Opt):cpu-count (Opt):load-balanceuuid (Req):action (Req):total-nvm-count (Opt):internal-virtual-network-ips (Opt):external-virtual-network-ips (Opt):memory (Opt):cpu-count (Opt):upgrade | upgradeversion (Req):uuid (Opt):upgrade-all-file-servers (Opt):
  • Userlevel 3
    Badge +17
    Hi dlink7,

    Thank you a lot for your answer

    By the way, I also need you explain to me a little bit in detail due to some specifical terminology

    1. I have already heard about the General Share and Home Share but what is top level directories (TLD) and how can I see the DSF tab of TLD?
    And... I know that the client data will be reside in a VG (General share). Am I able to verify which FSVM is responsible for that VG?
    With Home Share, the client data is spread among all of the VGs mounted by all FSVM, isn't it?

    2. What is ADRS?
    And the number of Home Share VGs depends on the number of FSVM, doesn't it? Because in my case, I have 4 FSVMs and the Home Share consists of up to 20 VGs so far

    4. Because I saw that the FSVM mounted the VGs via iSCSI so I wonder that whether FSVM mounted them via CVM? The VGs plays as Target and FSVM plays as Initiator in iSCSI client server model roles, right?

    I mean just like this: Client FSVM CVM VGs ?!
    I dont know exactly how FSVM could mount the VGs in your diagram

    5. By the way,
    I saw that the VG having 6 vDisk that has lot of capacity inside that almost exceed the physical capacity of cluster. Is that virtual capacity? and what metadata vDisk is 512GB and data disk 10TiB means to the true operation in AFS?

    6. Currently, I already created the File Server with 1TiB and created the General Share with the same capacity but when I access/map this Share as a drive to client, this drive is 40TiB due to the size of vDisk data right?
    Userlevel 4
    Badge +21
    TLD - is just the first level of folders in the share. If you right click on them there is a DFS tab which show where they are located (FSVM)ADRS - Acropolis Distributed Resource ScheduleIn 5.0 - Home share starts off with 15 Volume groups. It's not directly related to the number of FSVMSthe vdisks in the volume groups act as the target. . FSVM uses iSCSI to get access to the vdisks in the volume groups. Indivdual vdisks can use all of the CVMAll vdisks are thin provisionedneed to check the behaviour for number 6.0 in 5.0
    Userlevel 4
    Badge +21
    you are correct, confirmed it's 5 VGS per 1 FSVM
    Userlevel 4
    Badge +21
    your are correct for the home share, 5 VGS per FSVM.
    Userlevel 3
    Badge +17
    Hi dlink7

    Thanks a lot for your reply

    By the way, you are talking about the thin provisioning as all of the vdisk are recently using thin provisioning so that is the reason why we have 40TiB in data vdisks?

    This 40TiB number is unchanged regardless the capacity of FileServer as well as the Share?! May we delete the vDisk or add more vDisk in the automatically created VG?

    I am sorry, I still cannot get what you told me about the relationship between FSVM, CVM and DSF (vDisks). In this case, CVM acts as a bridge between FSVM and DSF data communication and/or iSCSI sessions? Could you please explain a little bit about it?

    Thanks in advance
    Userlevel 4
    Badge +21
    The backing disks were the user is sitting is on 1 of the 40 TB volume groups.

    You can't delete the volume groups. It's a NO TOUCH ZONE :-)

    FSVM is no different than any VM using Acropolis Block Services but it was automated failover - just read the nutanixbible on ABS - http://nutanixbible.com/
    you can starting at the below section
    Block Services
    The Acropolis Block Services (ABS) feature exposes back-end DSF storage to external consumers (guest OS, physical hosts, containers, etc.) via iSCSI.
    Userlevel 3
    Badge +17
    Hi dlink7

    I think I already got what you said to me and provide to me by lots of valuable information, knowledge

    Thank you so much for supporting me in this thread
    Userlevel 3
    Badge +17
    Hi dlink7

    Sorry for bring this up again cuz I am having one more question related to AFS.

    Everytime I create a File server cluster, one Protection Domain is automatically created.

    Is this PD responsible for WPV functions in FS Share in which the file/folder can be restored to previous state thanks to snapshot created by PD?

    If not, could you please explain to me the meaning of PD existence?

    Thanks
    Userlevel 4
    Badge +17
    Folks

    what license required for AFS ?
    as AFS running using starter license
    what happened if keep using AFS with starter license ?
    Userlevel 4
    Badge +21
    There is a stand alone license or you can buy ultilmate.

    It will continue to work with starter but you probably won't get support if you call in.
    Userlevel 4
    Badge +17
    dlink7 could you share link regarding AFS licensing ?

    Reply