[OpenAFS-devel] Re: [OpenAFS] namei interface lockf buggy on Solaris (and probably HP-UX and AIX)

[Robert Banz]

> Whew, you ain't kidding.  When I looked at that, I believe a lot of it
> was the link table.  I have been idly thinking about simply removing
> most of those fsync() calls, or collapsing a whole bunch of  
> them ... it
> would probably speed up operations like volume clones a whole lot.  A
> few thought experiments made me think that perhaps the consequences of
> an incorrect link count aren't so catastrophic that that salvager
> couldn't easily recover from it ... but AFS has fooled me before, so
> I'm not convinced of that yet :-)

Well, the OS should be syncing dirty buffers to disk in some sort of  
reasonable timeframe anyhow, so writing the code to batch the fsync's  
would simply be us doing the OS's work for it.

I built a fileserver/volserver that's fsync() less (as in never) and  
have been doing some basic testing.  First, volrestore operations  
seem much faster.  Also did a power-off test on a pretty active  
volume, and salvager didn't seem to have much of a problem with  
it...  Going to play around more with it and see what happens...

-rob