On Jun 29, 2011, at 1:01 PM, Gregory Farnum wrote: > Ah. Each directory has a lot of different locks protecting different pieces of state. Here, the only write lock that is being held for the creates in dir b is the nestlock, which protects the rstats (recursive stats). Generally that lock shouldn't block creates though. How long does your test take to run? > Since the lock says it's flushing I wonder if there's something else going on that's hurting your OSD performance and stalling it out. > > I'll see if I can reproduce it locally. I reproduced this locally and talked it over with Sage. The problem was just an unwise strategy for dealing with cache coherency that we applied a fairly simple fix for. It's in master on commit 310032ee8128f6417ac302c0f5ecd27c691cbcc7, which you should be able to cherry-pick if you wish. :) -Greg-- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
