On Thu, Apr 19, 2012 at 12:04 PM, Noah Watkins <jayhawk@xxxxxxxxxxx> wrote: > I've had a lot of success using rados_exec(), but I need to graduate the implementation to use IoCtx::aio_exec for increased concurrency. I'm a little confused about how aio_exec should be used: > > Here is aio_exec: > > int aio_exec(const std::string& oid, AioCompletion *c, const char *cls, const char *method, bufferlist& inbl, bufferlist *outbl); > > I'd like to have many outstanding competitions, one for each call of aio_exec. Once @c has completed, I'd like to have access to the @outbl bufferlist. So, is the intended use here to allocate an empty bufferlist for each submitted completion? Yes, you'll need to have a bufferlist for each request sent if your exec is expected to return data. Note that only read operations should return data. > > In rgw/rgw_rados.cc a use is: > > AioCompletion *c = librados::Rados::aio_create_completion(NULL, NULL, NULL); > r = io_ctx.aio_exec(oid, c, "rgw", "dir_suggest_changes", in, NULL); > c->release(); > > To clarify, are the semantics here that an asynchronous aio_exec is made without checking for completion or return values. Right. We send the "dir_suggest_changes" request and dropping the reference, so that we can't get the return value once it's done. Yehuda -- 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
