[OpenAFS] Re: bonnie++ on OpenAFS

Simon Wilkinson sxw@inf.ed.ac.uk
Tue, 23 Nov 2010 11:23:03 +0000

On 23 Nov 2010, at 11:02, Hartmut Reuter <reuter@rzg.mpg.de> wrote:
> The problem here ist that afs_DoPartialWrite is called with each write. No=
rmally it gets out without doing anything, but if the percentage of dirty ch=
unks is to high it triggers a background store.

On master, at last DoPartialWrite does an immediate store - the only place w=
e can do a background write is in response to a normal close request.

In any case, this problem arises regardless of how we're storing the file. T=
he issue is that our cache eviction strategy picks the most recently accesse=
d chunk to evict, and then we dirty that chunk again immediately after we've=
 flushed it.

We need a better solution to cache eviction. The problem is that, until very=
 recently, we didn't have the means for one process to successfully flush fi=
les written by a different process.