Jump to content
Updated Privacy Statement
  • 0

Is it possible to force NFS storage into asynchronous mount


Matthew Weiner

Question

I'm using Hypervisor 8.2 with an NFS mount on Dell storage, currently an FS8600 but likely moving to a PowerStore 1200T at some point in the future.  Being in synchronous mode the performance, well, not so good.  The problem I have is that according to the documentation Hypervisor depends on the storage advertising the mount as async, but on my storage there's no way to configure it to do so.  According to Dell it's a client thing, so I have a bit of a catch-22, the server says the client should do it, the client says the server should. 

 

Is there any way with mount options, like via the CLI, to just tell Hypervisor to mount the NFS volume async?  I have UPSes, standby generators, battery backed cache, all that stuff, so the infentesimal risk of data loss vs. a potential significant performance gain is worth it.

Link to comment

6 answers to this question

Recommended Posts

  • 0

Well that's unfortunate.  Dell confirmed there is no way to do this on the FS8600 family and from the documentation I found on the PowerStore that is the same.  Performance is pretty bad in sync mode, at least on the FS8600s, and the experiments I've done with FreeNAS, etc. it's not too much better.  I'll have to try to lean on Dell to see if they'll add async mount mode to the PowerStore's capabilities.

Link to comment
  • 0
12 hours ago, Kyle Peterson said:

We use a netapp A250 and netapp wants nfs to be async as well but citrix defaults to sync.

No, no it doesn't. The setting is server side in the NFS storage, there is no setting that Citrix Hypervisor/XenServer can set to force Asynchronous I/O to the Filer.

Link to comment
  • 0

The async mode supported by the Linux Kernel NFS server technically violates the NFS Specification (and is documented as such) so it's not entirely surprising that vendors do not supoprt it in their expensive hardware filers as it reduces the guarantees for data integrity.

 

XenServer quite rightly asserts that I/O must be confirmed completed by the remote storage before those I/O operations are reported as completed to the VM guests and if this results in the filer having to complete expensive operations then the fault lies in the filer.

 

For the NetApp in particular this should not be the case as all write operations go first to non-volatile memory (and are thus safe) and then are flushed to the backing storage in the background.

Link to comment

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...