On 10/14/11 11:28 AM, "Roger Evans" <[email protected]> wrote:
>What a hassle! > >I tested some today, and found that indeed, it takes some time before >changes are seen on the R/O disk. But if I do a sync after my database >backup (on the RW linux), then umount/mount pm the RO, > the changes are immediately visible. Why should I have to do a >detach/attach? Because the two systems are in completely different machines, and are unconscious of each other and particularly of cached data. Detach/attach is the only time Linux forces a re-read of the data actually on the disk for the R/O systems, so you need to sync on the R/W system, then detach/attach on the R/O systems to pick up the actual changes. >I'm still using ext2 because I'm >sceptical about using XFS, Isn't it a journaling filesystem? Yes, but it's smart enough to allow journals to be turned off and on with commands. Journals don't really matter for R/O filesystems (they only matter with R/W), and the prefetch and buffering algorithms in XFS are a better suited for RO configurations (they work really well for systems that are based on flash, which is a close analogue to the one writer/multiple reader configuration that Agblad was talking about). ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For more information on Linux on System z, visit http://wiki.linuxvm.org/
