Heavy SMB signing pressure causes high SMB3 latency from Network CPU domain
Applies to
- Versions of ONTAP prior to 9.7
- SMB3
- SMB2
Issue
- High SMB latency from Unified Manager, Grafana.
- In the latency breakdown section, the bulk of the latency is from
Network
orCPU N-blade
, depending on the tools that are used to monitor performance
- In the latency breakdown section, the bulk of the latency is from
- The impacted Vservers have a large amount of SMB3 or SMB3.1 connections
cluster1::> set -c off advanced cluster1::*> statistics start -node node1 -object cifs -counter total_smb3_1_connections_count|total_smb3_connections_count cluster1::*> statistics show -sample-id sample_1 Object: cifs Instance: testsvm Start-time: 11/22/2021 15:39:20 End-time: 11/22/2021 15:42:00 Elapsed-time: 160s Scope: testsvm Counter Value -------------------------------- -------------------------------- total_smb3_1_connections_count 1031 total_smb3_connections_count 1254
- The impacted Vservers have signed sessions, as well as a significant amount of data being signed per second
cluster1::*> statistics start -node node1 -object smb2 -counter bytes_signed_rate|signed_sessions
cluster1::*> statistics show -sample-id sample_2
Object: smb2
Instance: testsvm
Start-time: 11/22/2021 15:48:17
End-time: 11/22/2021 15:49:05
Elapsed-time: 48s
Scope: testsvm
Counter Value
-------------------------------- --------------------------------
bytes_signed_rate 10240000B
signed_sessions 128
- Network CPU is not necessarily saturated
- Avg CPU might NOT be high
Nwk_Exmpt
CPU domain from systat -M
- Other protocols might be impacted as well, such as NFS or iSCSI. As they have to complete for the same network threads with SMB signing workload