0

I have some (6) boxes that have an NFS share mounted on which they write files. All run ubuntu 14.01 and the NFS server is a standard nfs-server from ubuntu packages. This works well.

However, when renaming files on the share by one system, it can take around half a minute before the other systems see the rename. Calling sync on the client after the rename doesn't help. Is this delayed write caching in action?

glormph
  • 115
  • 4

1 Answers1

1

Set actimeo to a lower value in your fstab file and remount.

Xavier Lucas
  • 12,815
  • 2
  • 44
  • 50
  • That helped a little, files appeared faster, but even at actimeo=0 it was still not fast enough. Maybe I should switch to another shared FS. – glormph Jan 23 '15 at 10:10
  • @glormph You did set this option on both systems right ? This should lower the behaviour you see to one second if `actimeo=1`. If you don't want any file attribute caching client-side set the `noac` option. Be aware that it makes writes synchronous thus brings a performance penalty. Are you running a caching daemon like FS-Cache at the same time ? – Xavier Lucas Jan 23 '15 at 13:39
  • Yes, on both systems. Setting actimeo to 1 seems to work perfectly well with files in a test script though, they get renamed almost instantly, so thanks for the tip and I'll accept. But confused, it's like extra caching happens just for that process. Will maybe solve it very badly with a call to sleep or something. – glormph Jan 26 '15 at 14:20