Re: RFC: default group_isolation to 1, remove option

[Vivek Goyal <vgoyal@redhat.com>]

On Mon, Mar 07, 2011 at 03:24:32PM -0500, Vivek Goyal wrote:

[..]
> > 
> > Or not get rid of limits completely, but do a lot more relaxed
> > accounting at the queue level still. That will not require any
> > additional tracking of io contexts etc, but still impose some limit on
> > the number of queued IOs.
> 
> A lot more relaxed limit accounting should help a bit but it after a 
> while it might happen that slow movers eat up lots of request descriptors
> and making not much of progress.
> 
> Long back I had implemented this additional notion of q->nr_group_requests
> where we defined per group number of requests allowed submitter will
> be put to sleep. I also extended it to also export per bdi per group
> congestion notion. So a flusher thread can look at the page and cgroup
> of the page and determine if respective cgroup is congested or not. If
> cgroup is congested, flusher thread can move to next inode so that it
> is not put to sleep behind a slow mover.
> 
> Completely limitless queueu will solve the problem completely. But I guess
> then we can get that complain back that flusher thread submitted too much
> of IO to device.

Also wanted to add that currently blk-throttling code implements limitless
queuing of bio. The reason I did not enforce the limit yet because of
same reason that I will run into issues with async WRITES and flusher
thread.

So once we have figured out what't the right thing to do here, I can
implement similar solution for throttling too.

One side affect of limitless bio queueing is an AIO process can queue up lots
of bios in a group and if one tries to kill the process, it waits for all the
IOs to finish and can take up a very long time depending on throttling limits
of the group.

Thanks
Vivek