Hi Brian, > -----Original Message----- > From: ceph-users [mailto:[email protected]] On Behalf Of > Brian Kroth > Sent: 23 October 2015 21:31 > To: [email protected] > Subject: [ceph-users] cache tier write-back upper bound? > > Hi, I'm wondering when using a cache pool tier if there's an upper bound on > when something written to the cache is flushed back to the backing pool? > Something like a cache_max_flush_age setting? Basically I'm wondering if I > have the unfortunate case of all of the SSD replicas for a cache pool object > all > go at once, how far behind is the backing pool object from the latest data?
There is nothing like this currently, data will reside until it gets forced out by newer data entering the cache pool. I understand your concern about losing data, but in that case I would make sure that I am using sufficient replicas in the SSD tier to protect against it. Anyway I suspect that even if you only had a couple of GB left in the cache tier, a loss of the cache tier would still likely leave devastating holes all over your data. > > Also, am I reading things correctly that if you wanted to turn the write-back > mode into something close to a write-through (though not exactly), you'd do > something like the following? > > # ceph osd pool set cachepool cache_target_dirty_ratio 0.00 # ceph osd pool > set cachepool cache_min_flush_age 0 Possibly, I'm not sure how these values effect the cache agent when set to 0, you might want to increase them to just slightly above 0. Another thing to keep in mind is that a lot of the current performance problems with using cache tiering are related mainly to the amount of data movement between tiers. Generally the best approach has been to try and minimise this data movement by making all the data fit into the cache tier, but YMMV. > > That should still ack the client as soon as the replicas were confirmed on the > cachepool layer, but then immediately let the background flusher start > writing the updates to the backing pool, all while still leaving the object > available for further updates from clients, correct? Or does the background > flusher need to lock the object while it writes it to the backing pool, thus > stalling further client updates to to the object until that completes? > > I'm guessing that setting cache_target_dirty_ratio to 0 and > cache_min_flush_age to N, still wouldn't quite implement > cache_max_flush_age since if the object is continually getting updated, then > that timer is continually getting reset, so it never becomes a candidate to > get > updated in the backing store, right? > > Thanks, > Brian > _______________________________________________ > ceph-users mailing list > [email protected] > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com Nick Fisk Technical Support Engineer System Professional Ltd tel: 01825 830000 mob: 07711377522 fax: 01825 830001 mail: [email protected] web: www.sys-pro.co.uk<http://www.sys-pro.co.uk> IT SUPPORT SERVICES | VIRTUALISATION | STORAGE | BACKUP AND DR | IT CONSULTING Registered Office: Wilderness Barns, Wilderness Lane, Hadlow Down, East Sussex, TN22 4HU Registered in England and Wales. Company Number: 04754200 Confidentiality: This e-mail and its attachments are intended for the above named only and may be confidential. If they have come to you in error you must take no action based on them, nor must you copy or show them to anyone; please reply to this e-mail and highlight the error. Security Warning: Please note that this e-mail has been created in the knowledge that Internet e-mail is not a 100% secure communications medium. We advise that you understand and observe this lack of security when e-mailing us. Viruses: Although we have taken steps to ensure that this e-mail and attachments are free from any virus, we advise that in keeping with good computing practice the recipient should ensure they are actually virus free. Any views expressed in this e-mail message are those of the individual and not necessarily those of the company or any of its subsidiaries. _______________________________________________ ceph-users mailing list [email protected] http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
