VMware Solutions Discussions

Quest: what's the expect behavior about GOS timeout and NFS timeout.

chao
2,633 Views

Hi Gurus

VSC will use the following NFS setting:

NFS.HeartbeatMaxFailures......10

NFS.HeartbeatFrequency........12

NFS.HeartbeatTimeout..........5

that means ESXi will mark NFS datastore as failure after 125s

For GOS timeout, According the VSC 4.1 manual:

There are two scripts for each operating system:

  • A 60-second script
  • A 190-second script

I was a little confused,  What's the relation btw EXSi NFS timeout and GOS timeout?  

if use 60-second script,  if NFS datastore has no response over 60s but less than 125s,  what's the GOS happen?  Does it report SCSI error and hung the application?

if use 90-second script,  if NFS datastore has no response over 125s and was marked as failure,  how does GOS 190s timout work?

Thanks and Best Regards!

TC

2 REPLIES 2

schrie
2,633 Views

The GOS timeout scripts are available to ensure that the disk timeout settings within a VM can "wait" the same amount of time for I/O to resume in the case of any break in communication on the underlying host - for example, during a takeover or giveback operation on the controller.  While the NFS settings will ensure that the physical ESX server waits an appropriate amount of time before disconnecting the storage - the GOS scripts will do the same for the OS running inside of the VM, for each SCSI disk.  The different times are available depending on which version of ESX/VSC/ONTAP/GOS is being run, and typically the 190s is only required for older combinations.

chao
2,633 Views

Hi schrie

I agree.  I just wonder why VSC 4.1 provide 60-seconds scripts and sounds recommend to use this for any new VMs.  (see below)

in most cases, the recommended value is 60 seconds. Knowledge base article 3013622, which is online at kb.netapp.com/support/index?page=content&id=3013622, contains information you can use when deciding which timeout value to use.

I also check this KB and this KB subject is  "What are the guest OS tunings needed for a VMware SAN deployment?"

I think we have to face same scenario when use  NFS datastore.  so how 60-seconds GOS timeout setting work with ESXi 125 seconds NFS setting?   the reason that I raised this case is for how to define cDOT volume move cut-over windows.  by default, the volume move cut-over window is 45s, but in some case, 45 seconds is not enough.  according the ESXi NFS setting, we could enlarger it to <125 second  but not sure how it impact GOS.

BR

TC

Public