Higher than expected SSD latency due to missing 12Gb SAS cables

Last updated
Save as PDF
Share
1. Share
2. Tweet
3. Share

Views:: 1,097

Visibility:: Public

Votes:: 0

Category:: aff-series

Specialty:: perf

Last Updated:

Applies to

ONTAP 9
Data ONTAP 8
AFF
12Gb SAS connectors

Issue

Under certain rare circumstances, All Flash FAS (AFF) storage controllers may experience higher than expected latency numbers at the disk layer.
The larger than expected latency can be observed in the statit command output or qos statistics volume latency show command output.

Example:

qos statistics volume latency show command output showing disk latency:

cluster::> qos statistics volume latency show Workload ID Latency Network Cluster Data Disk Qos Max Qos Min NVRAM --------------- ------ -------- -------- -------- -------- -------- -------- ---------- ---------- -total- 10.35ms 1.35ms 0ms 0us 9ms 0ms 0ms 0ms vs1vol0 111 17.23ms 0us 0ms 603.00us 16.63ms 0ms 0ms 0ms vol1 1234 17.76ms 0ms 0ms 150.00us 17.61ms 0ms 0ms 0ms vol2 999 4.24ms 0us 0ms 190.00us 4.05ms 0ms 0ms 0ms

The io_pending and the io_queued values are high
The SSD disks show high latency and utilization in statit

Cluster::> set -privilege diag

Warning: These diagnostic commands are for use by NetApp personnel only.
Do you want to continue? {y|n}: y

Cluster::*> statistics start -object disk -counter "io_queued|io_pending"
Statistics collection is being started for sample-id: sample_148


Cluster::*> statistics show -filter "io_queued>10|io_pending>10"


Object: disk
Instance: 0d.23.13
Start-time: 12/5/2022 16:48:26
End-time: 12/5/2022 16:51:58
Elapsed-time: 212s
Scope: node1
Number of Constituents: 1 (complete_aggregation)
    Counter                                                     Value
    -------------------------------- --------------------------------
    io_queued                                                      19
    io_pending                                                      8

...

Cluster::*> node run -node node1 -command statit -b
Cluster::*> node run -node node1 -command statit -e
                       Disk Statistics (per second)  
        ut% is the percent of time the disk was busy.  
        xfers is the number of data-transfer commands issued per second.  
        xfers = ureads + writes + cpreads + greads + gwrites  
        chain is the average number of 4K blocks per command.  
        usecs is the average disk round-trip time per 4K block.  
  disk            ut%  xfers   ureads--chain-usecs writes--chain-usecs cpreads-chain-usecs greads--chain-usecs gwrites--chain-usecs
/data_aggr1/plex0/rg0:                                                                  
0c.01.2           80 2510.99  1525.62   5.95   763 389.58  11.11  2682 595.79   5.22  1318   0.00   ....     .   0.00   ....     .
0c.01.3           81 2502.32  1510.75   5.96   758 393.43  11.27  2706 598.14   5.15  1348   0.00   ....     .   0.00   ....     .
0c.01.4           81 2518.91  1528.01   5.93   762 392.18  11.45  2661 598.72   5.12  1364   0.00   ....     .   0.00   ....     .
...