On Thu, 15 Dec 2016, Igor Fedotov wrote:
> Yeah, 0.9 is much better than 0.5.
> Another/additional option to consider might be forcing clients that need
> caching to request that explicitly. E.g. by specifying a need_cache flag for
> reads. And doesn't do it by default.
This is pretty close to what happens now. There are
bluestore_defautl_buffered_{read,write} options that default to true for
read and false for write.
sage
>
>
>
>
> Sent from Samsung tablet.
>
>
> -------- Original message --------
> From: Sage Weil
> Date:2016/12/15 21:11 (GMT+03:00)
> To: Igor Fedotov
> Cc: Mark Nelson ,ceph-devel
> Subject: Re: Metadata vs UserData caching in BlueStore
>
> On Thu, 15 Dec 2016, Igor Fedotov wrote:
> > On 15.12.2016 20:34, Mark Nelson wrote:
> > >
> > >
> > > On 12/15/2016 11:24 AM, Igor Fedotov wrote:
> > > > Sage et al,
> > > >
> > > > shouldn't we have bluestore_cache_metadata_ratio set to 1 by default?
> To
> > > > concentrate on metadata caching..
> > >
> > > I would contend that it should be high (we've already bumped it up, I'd
> go
> > > higher still), but isn't 1 a little excessive? Ultimately this should
> all be
> > > automatic based on heuristics imho.
> > >
> > Well, Mark - what's the use case for user data caching? It means that user
> > either retrieves just written data or reads them repeatedly or in an
> unaligned
> > manner. Or there are multiple users working with the same object.
> > Any of that is IMHO a corner case... That can be handled by using that
> ratio
> > setting if needed.
>
> The main one is RGW. The gateways don't do their own caching, and even
> if they did you often have several of them. CephFS multiclient
> scenarios can exercise the cache too.
>
> I agree, though. How about .8 or .9?
>
> sage
>
>
