Forum Discussion

AdN's avatar
AdN
Level 4
2 years ago

Slow NBD performance when backing up VMs

Hello,

We are using VMware vSphere 7 running on Cisco UCS blade system. VMDK data storage is NFS on a Netapp AFF A400 storage system. NetBackup 10.1 media server is running on RHEL8 in a VM and has BasicDisk NFS mounted volumes from a Netapp FAS 2720 storage system. From there backup image duplication is done off-site to a HP media server with StorageTek LTO5 tape library. All network uplinks are minimum 2 x 10 Gb/s and MTU 9000.

We used to have classic NetBackup via agent on VMware guest OS and switched now to VMware direct backup using NBD over LAN. We don't have a SAN. NetBackup VMware policies are configured with Java admin console. 

Backup speeds from VMware to disk are just around 10'000-20'000 KB/s. Netapp storage system and network is not the bottleneck as direct copy jobs are way faster. 

Where could the bottleneck be? What am I missing?

Thanks in advance.

Regards,
Bernd

Linux
  • AdN's avatar
    AdN
    2 years ago

    Found this one here ...
    https://kb.vmware.com/s/article/83401 
    Maybe the Netapp FAS 2720 is not the fastest for write operations as NFS BasicDisk backup target ...

    fas2720::*> statistics disk show -sort-key operation_latency

fas2720 : 7/3/2023 08:14:48

                Busy Total Read Write    Read Write *Latency
 Disk      Node  (%)   Ops  Ops   Ops   (Bps) (Bps)     (us)
----- --------- ---- ----- ---- ----- ------- ----- --------
1.3.5 adncl3-01   43    52   51     0 1773568  1024    24856
1.2.11
      adncl3-01   50    52   52     0 1957888  2048    21840
1.2.7 adncl3-01   48    57   57     0 2040832     0    21415
1.4.9 adncl3-01   52    65   65     0 2387968  1024    21167
1.5.7 adncl3-01   47    50   50     0 2055168     0    20383
1.3.7 adncl3-01   51    68   68     0 2185216     0    19792
1.3.9 adncl3-01   45    57   57     0 2135040     0    19122
1.3.11
      adncl3-01   44    52   52     0 2073600     0    19058
1.1.11
      adncl3-01   44    57   57     0 2018304 21504    18197
1.4.7 adncl3-01   43    54   54     0 2242560  3072    18115
1.5.9 adncl3-01   44    60   59     0 2264064  2048    16991
1.2.9 adncl3-01   43    62   62     0 2221056  3072    16816
1.0.3 adncl3-01   24    72   69     2 2379776 175104
                                                       15037
1.1.9 adncl3-01   45    69   68     0 2337792  1024    14815
1.5.1 adncl3-01   33    58   57     0 1985536  4096    13554
1.1.5 adncl3-01   31    54   53     0 1774592  2048    13319
1.4.1 adncl3-01   34    58   58     0 1604608  3072    13197
1.3.3 adncl3-01   32    54   54     0 1701888 10240    13128
1.2.5 adncl3-01   32    53   53     0 1820672     0    12883
1.4.5 adncl3-01   33    63   62     0 2308096 10240    12650
1.5.5 adncl3-01   32    52   52     0 2238464  1024    12607
1.4.3 adncl3-01   32    53   53     0 1920000  2048    12573
1.5.3 adncl3-01   35    59   59     0 2415616  2048    12140
1.2.3 adncl3-01   31    53   53     0 1628160     0    11767
1.1.3 adncl3-01   30    55   55     0 2255872     0    11027

     Compared to Netapp AFF A400 all-flash system for VMware VMDK disk storage.

    affa400::*> statistics disk show -sort-key operation_latency

affa400 : 7/3/2023 08:32:56

                      Busy Total Read Write     Read    Write *Latency
 Disk            Node  (%)   Ops  Ops   Ops    (Bps)    (Bps)     (us)
----- --------------- ---- ----- ---- ----- -------- -------- --------
1.1.0 Multiple_Values    0   387  342    44 11252736 10230784      152
1.0.4 Multiple_Values    0   431  373    58 12254208 11289600      150
1.1.21
      Multiple_Values    0   433  377    56 12266496 11183104      141
1.0.17
            adncl2-02    0   427  294   132  6013952  6899712      136
1.0.16
      Multiple_Values    0   788  654   134 11297792  6376448      124
1.0.0 Multiple_Values    0   404  357    46 11068416 10337280      124
1.0.19
      Multiple_Values    0   809  667   142 10535936  6147072      122
1.0.14
      Multiple_Values    0   797  648   149 11408384  6798336      122
1.1.3 Multiple_Values    0   422  287   134  5437440  5969920      120
1.0.11
      Multiple_Values    0   834  678   156 12203008  6723584      118
1.0.1 Multiple_Values    0   799  647   152 11014144  5924864      118
1.1.5 Multiple_Values    0   772  639   133 11672576  5780480      117
1.1.18
      Multiple_Values    0   445  308   137  5679104  5788672      117
1.0.3 Multiple_Values    0   791  655   135 12166144  5995520      117
1.0.12
      Multiple_Values    0   803  662   141 11409408  6824960      117
1.0.9 Multiple_Values    0   726  606   119 11119616  6731776      116
1.1.1 Multiple_Values    0   815  667   148 10887168  6126592      112
1.1.20
      Multiple_Values    0   742  613   129 10509312  5449728      111
1.1.19
      Multiple_Values    0   795  648   147 11213824  5847040      111
1.0.8 Multiple_Values    0   814  659   154 10331136  5297152      111
1.1.23
      Multiple_Values    0   756  638   118 11285504  6646784      110
1.1.2 Multiple_Values    0   780  633   147 10660864  6048768      110
1.1.4 Multiple_Values    0   795  638   156 11356160  6163456      109
1.1.22
      Multiple_Values    0   759  627   132 10919936  6013952      109
1.0.7 Multiple_Values    0   782  632   149 11244544  6273024      105


     

  • AdN's avatar
    AdN
    Level 4
    2 years ago

    Ah, HotAdd also works with NFS datastores. They just need all to be mounted on the ESXi server where the media server virtual machine is running. That is around 10x faster than NBD.

  • AdN's avatar
    AdN
    Level 4
    2 years ago

    Just had an incident over weekend where the backup media server could not unmount the virtual disk files of the backup client. So the snapshot consolidation of the client filed with timeout errors.

    So, the HotAdd transport method is 10x faster but evil. :(

  • AdN's avatar
    AdN
    Level 4
    2 years ago

    Any ideas how NBD speed can be improved? Netapp NFS storage and network is not the bottleneck and also not CPU/memory of the media server VM.

    I looked at Replication Director that it can backup VMs using Netapp Snapshots/SnapMirror, but I could not see that the backup images then can be staged to tapes.

  • AdN's avatar
    AdN
    Level 4
    2 years ago

    Found this one here ...
    https://kb.vmware.com/s/article/83401 
    Maybe the Netapp FAS 2720 is not the fastest for write operations as NFS BasicDisk backup target ...

    fas2720::*> statistics disk show -sort-key operation_latency

fas2720 : 7/3/2023 08:14:48

                Busy Total Read Write    Read Write *Latency
 Disk      Node  (%)   Ops  Ops   Ops   (Bps) (Bps)     (us)
----- --------- ---- ----- ---- ----- ------- ----- --------
1.3.5 adncl3-01   43    52   51     0 1773568  1024    24856
1.2.11
      adncl3-01   50    52   52     0 1957888  2048    21840
1.2.7 adncl3-01   48    57   57     0 2040832     0    21415
1.4.9 adncl3-01   52    65   65     0 2387968  1024    21167
1.5.7 adncl3-01   47    50   50     0 2055168     0    20383
1.3.7 adncl3-01   51    68   68     0 2185216     0    19792
1.3.9 adncl3-01   45    57   57     0 2135040     0    19122
1.3.11
      adncl3-01   44    52   52     0 2073600     0    19058
1.1.11
      adncl3-01   44    57   57     0 2018304 21504    18197
1.4.7 adncl3-01   43    54   54     0 2242560  3072    18115
1.5.9 adncl3-01   44    60   59     0 2264064  2048    16991
1.2.9 adncl3-01   43    62   62     0 2221056  3072    16816
1.0.3 adncl3-01   24    72   69     2 2379776 175104
                                                       15037
1.1.9 adncl3-01   45    69   68     0 2337792  1024    14815
1.5.1 adncl3-01   33    58   57     0 1985536  4096    13554
1.1.5 adncl3-01   31    54   53     0 1774592  2048    13319
1.4.1 adncl3-01   34    58   58     0 1604608  3072    13197
1.3.3 adncl3-01   32    54   54     0 1701888 10240    13128
1.2.5 adncl3-01   32    53   53     0 1820672     0    12883
1.4.5 adncl3-01   33    63   62     0 2308096 10240    12650
1.5.5 adncl3-01   32    52   52     0 2238464  1024    12607
1.4.3 adncl3-01   32    53   53     0 1920000  2048    12573
1.5.3 adncl3-01   35    59   59     0 2415616  2048    12140
1.2.3 adncl3-01   31    53   53     0 1628160     0    11767
1.1.3 adncl3-01   30    55   55     0 2255872     0    11027

     Compared to Netapp AFF A400 all-flash system for VMware VMDK disk storage.

    affa400::*> statistics disk show -sort-key operation_latency

affa400 : 7/3/2023 08:32:56

                      Busy Total Read Write     Read    Write *Latency
 Disk            Node  (%)   Ops  Ops   Ops    (Bps)    (Bps)     (us)
----- --------------- ---- ----- ---- ----- -------- -------- --------
1.1.0 Multiple_Values    0   387  342    44 11252736 10230784      152
1.0.4 Multiple_Values    0   431  373    58 12254208 11289600      150
1.1.21
      Multiple_Values    0   433  377    56 12266496 11183104      141
1.0.17
            adncl2-02    0   427  294   132  6013952  6899712      136
1.0.16
      Multiple_Values    0   788  654   134 11297792  6376448      124
1.0.0 Multiple_Values    0   404  357    46 11068416 10337280      124
1.0.19
      Multiple_Values    0   809  667   142 10535936  6147072      122
1.0.14
      Multiple_Values    0   797  648   149 11408384  6798336      122
1.1.3 Multiple_Values    0   422  287   134  5437440  5969920      120
1.0.11
      Multiple_Values    0   834  678   156 12203008  6723584      118
1.0.1 Multiple_Values    0   799  647   152 11014144  5924864      118
1.1.5 Multiple_Values    0   772  639   133 11672576  5780480      117
1.1.18
      Multiple_Values    0   445  308   137  5679104  5788672      117
1.0.3 Multiple_Values    0   791  655   135 12166144  5995520      117
1.0.12
      Multiple_Values    0   803  662   141 11409408  6824960      117
1.0.9 Multiple_Values    0   726  606   119 11119616  6731776      116
1.1.1 Multiple_Values    0   815  667   148 10887168  6126592      112
1.1.20
      Multiple_Values    0   742  613   129 10509312  5449728      111
1.1.19
      Multiple_Values    0   795  648   147 11213824  5847040      111
1.0.8 Multiple_Values    0   814  659   154 10331136  5297152      111
1.1.23
      Multiple_Values    0   756  638   118 11285504  6646784      110
1.1.2 Multiple_Values    0   780  633   147 10660864  6048768      110
1.1.4 Multiple_Values    0   795  638   156 11356160  6163456      109
1.1.22
      Multiple_Values    0   759  627   132 10919936  6013952      109
1.0.7 Multiple_Values    0   782  632   149 11244544  6273024      105