On Wed, Jul 17, 2013 at 9:50 AM, Da Chun <[email protected]> wrote: > > Do you mean the "write barrier"? > So all ceph disk partitions are mounted with "barrier=1"?
Yes. Ceph requires very strong safety semantics in order to maintain consistency, and it's very careful to do so in a performant fashion (instead of just syncing all over the place), but if the consistency mechanisms in the kernel are broken/disabled/whatever then it's all for naught. -Greg Software Engineer #42 @ http://inktank.com | http://ceph.com > > ------------------ Original ------------------ > From: "Gregory Farnum"<[email protected]>; > Date: Wed, Jul 17, 2013 00:29 AM > To: "Da Chun"<[email protected]>; > Cc: "ceph-users"<[email protected]>; > Subject: Re: [ceph-users] Should the disk write cache be disabled? > > Just old kernels, as they didn't correctly provide all the barriers > and other ordering constraints necessary for the write cache to be > used safely. > -Greg > Software Engineer #42 @ http://inktank.com | http://ceph.com > > > On Tue, Jul 16, 2013 at 9:20 AM, Da Chun <[email protected]> wrote: >> In this doc, >> >> http://ceph.com/docs/master/rados/configuration/filesystem-recommendations/ >> >> It says, >> >> Ceph aims for data safety, which means that when the Ceph Client receives >> notice that data was written to a storage drive, that data was actually >> written to the storage drive. For old kernels (<2.6.33), disable the write >> cache if the journal is on a raw drive. Newer kernels should work fine. >> >> Use hdparm to disable write caching on the hard disk: >> >> sudo hdparm -W 0 /dev/hda >> >> Does it mean the disk write cache should be always disabled for ceph, or >> just when using the old kernel (<2.6.33)? >> >> Thanks for your time! >> >> >> _______________________________________________ >> ceph-users mailing list >> [email protected] >> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com >> _______________________________________________ ceph-users mailing list [email protected] http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
