tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: Serious WAPL performance problems

Reinoud Zandijk <> wrote:
> Hi!
> On Wed, Oct 24, 2012 at 12:07:21AM +0100, Mindaugas Rasiukevicius wrote:
> > "Easy to diagnose problems"?  Plain false.  The lock naming you are
> > talking about would give no *more* information than "it is a vnode
> > lock", and one can guess already that it is most likely the case here
> > (what a surprise!).
> Would it help if we would name the vnode locks or other duplicated locks
> with their number? Say waiting for lock on "v12451"?
> just my $0.02,

If you establish some association between that number and e.g. a vnode,
that might be a little bit more useful.  However, usually you need to see
what exactly LWPs are doing (e.g. the order in which they acquire multiple
locks and so on; for this we need backtraces).  Merely determining the LWPs
and their last-locked vnodes is not enough, unless the case happens to be
very simplistic.

Also, instead of vnode lock naming, the same what you described can be
achieved in a cheaper way by storing the last-locked-vnode in struct lwp.
At this point, it is worth storing more data and more useful one, e.g. a
chain of acquired locks by that LWP (and perhaps performing some deadlock
detection in LOCKDEBUG).  This would help quite more than pretty labels.

> Reinoud


Home | Main Index | Thread Index | Old Index