On Thu, Apr 17, 2014 at 3:03 PM, Manuel Arostegui <man...@tuenti.com> wrote:
> > > > 2014-04-17 11:11 GMT+02:00 Ajay Garg <ajaygargn...@gmail.com>: > > On Thu, Apr 17, 2014 at 2:28 PM, Reindl Harald <h.rei...@thelounge.net >> >wrote: >> >> > >> > >> > Am 17.04.2014 10:55, schrieb Ajay Garg: >> > > I do understand the meaning of Unix "sync" function. >> > > >> > > So, you mean to say that "flushing" and "syncing" are same, in the >> > context of MySQL? >> > >> > please get rid of top-posting and reply-all >> > >> > a flush without a sync is "dear operating system, may i ask you to write >> > that to disk if you find the time to do so" while a sync is "write that >> > to disk" without a nice asking >> > >> >> Thanks Reindl, that clears it up !! > > > Keep in mind that if you are using HW RAID controller (and a BBU) the > concept of write-back and write-through are important to have in mind too > whilst thinking about how and when your data is written to disk. > If you enable sync_binlog and trx_commit to 1, it might not necessarily be > committing on every transaction to disk, as it might depend on how you've > configured your controller. > > If you have write-through configured, the data will be written to disk and > cache at the same time (this has a performance drawback). > If you have write-back configured, the data will be written to cache (and > then to disk, but _not_ at the same time) so you might lose data if there's > a power failure. When you have write-back, at an OS level your data has > been sync'ed to disk even though this is not totally true. > write-back is better for performance, but you have to keep in mind that > you can lose data (which is not common, but you're not 100% safe). > > Hope this helps > Thanks Manuel for the note !! It certainly is something to keep in mind !! > Manuel. > > -- Regards, Ajay