On Thu, Jul 31, 2014 at 2:35 PM, David Graham <xtnega at gmail.com> wrote: > Question: I've not gone through a setup yet, just an interested lurker > reading and interpreting capabilities at this time. > > my understanding of Ceph journal is that one can use a partition or a file > on a Filesystem.... if i use a filesystem, i can have assured integrity if > the FS does this for me (ala BTRFS) > > If i choose to use a partition for the journal: > --does this get auto-formatted with a filesystem? (if so, what?) No. > --I assume no filesystem; but is there check-summing of all data within the > journal? Yes; all operations are checksummed. > --Is the performance of the journal synchronous or asynchronous? I'm not sure I quite understand this question, but it's fairly configurable. I believe by default it's doing async IO on the journal. > --is the performance limited by the performance of the journals queue depth > 1 capability? (much like ifs) The OSD will queue up lots of sequential operations on the journal; don't assume the disk will only have a queue depth of 1. > --is the journal write pattern random, or largely sequential? Sequential. There's a header that gets written occasionally, but that's it. > --if i understand; is it correct to say that *all* writes, no matter sync or > async from the client will go into the journal, correct? Correct. -Greg Software Engineer #42 @ http://inktank.com | http://ceph.com
 |