Nutanix Files - FSVM Max Connections Notification

  • 10 December 2019
  • 3 replies
  • 2718 views

Userlevel 3
Badge +4

Nutanix Files offers SMB shares and NFS exports for file serving use-cases in enterprise environments. Not all environments are equal and every solution has some recommendations to cater for high traffic environments, where hundreds or thousands of users might be accessing a single storage endpoint or a file share concurrently.

Nutanix AOS and Files offers the flexibility to scale as your requirements grow without down-time or disruption to existing operations.

Connection count to existing file shares can increase, where a Nutanix Files environment is being used by more users or applications than initially anticipated at the time of deployment. 

An increase in users or usage of the Nutanix Files cluster can trigger a high connection count alert as follows:

FileServerVmMaxConnectionsNotification (A160039)

The "Maximum Connection limit" alert is generated when the Files cluster has one or more FSVM nodes under extensive load. This alert recommends running *’Performance optimisation'. This also indicates that there is a high connection count on FSVMs. 

If this alert is triggered, following Nutanix KBs, specifies the steps which should be followed to triage the connection load on File Server VM’s which can be above the recommended system limits.

Nutanix KB : FileServerVmMaxConnectionsNotification

Nutanix Files 3.6 Administration Guide : Logging into a FSVM

Nutanix Files Performance Optimisation

Nutanix Files Sizing Guide : File Storage use case


3 replies

Userlevel 2
Badge +4

What happens when “Alert A160080: CLUSTERNAME: Critical : For file server: FILESERVER share SOMESHARE is using 129 pct fsvm connection limit.” is raised?

The SMB share in question had users mapped to it, but wasn’t the share which was ‘causing’ the issue (which was that multiple robocopy connections with /MT:128 were being used on a 3 node FSVM cluster with 12GB RAM/4 cores only -base config)

 

Typically in the case of an SMB client saw this during some file copies:

The operating system returned the error '59(An unexpected network error occurred.)'

 

Overall the FSVM did great (IMHO) handling 100’s of connections without any issues, i.e. we were pushing it very hard with minimal resources, so this was almost ‘expected’… 

I would imagine that under load the next attempted session/connection pair is what gets dropped/rejected, right (hence the alert)?

Userlevel 3
Badge +4

What happens when “Alert A160080: CLUSTERNAME: Critical : For file server: FILESERVER share SOMESHARE is using 129 pct fsvm connection limit.” is raised?

The SMB share in question had users mapped to it, but wasn’t the share which was ‘causing’ the issue (which was that multiple robocopy connections with /MT:128 were being used on a 3 node FSVM cluster with 12GB RAM/4 cores only -base config)

 

Typically in the case of an SMB client saw this during some file copies:

The operating system returned the error '59(An unexpected network error occurred.)'

 

Overall the FSVM did great (IMHO) handling 100’s of connections without any issues, i.e. we were pushing it very hard with minimal resources, so this was almost ‘expected’… 

I would imagine that under load the next attempted session/connection pair is what gets dropped/rejected, right (hence the alert)?

 

Hi @DavidN thank you for your comment.

Not sure, if I understand your question correctly. 

so you were doing a robocopy on a files cluster and you got an alert for maxconnection alert.

it depends what type of share you were writing to, Nutanix Files have two types of shares:

Exports and Shares (Files TechNote)

There are two types of SMB shares and NFS exports.
SMB shares:

1. Distributed (previously called home).
2. Standard (previously called general).

NFS exports:
1. Distributed (previously called sharded).
2. Standard (previously called nonsharded).

You can check the connection count by logging in to a cvm and then type:

afs (to get to the afs prompt)
<afs> misc.get_leader (to get the minerva leader)

SSH to the minerva leader vm and look into the minerva_nvm.log
there you can search for “connection counts” to see which Client IP address is creating x number of connections to the FSVMs. Also, “MaxConnectionsNotification” can be searched as well in the log file to determine if the Cluster is complaining about a certain FSVM.

Hope this helps

BR

Userlevel 2
Badge +4

I think we’re hitting another issue - See ticket 00691243 

Reply