Hello Richard, Thursday, July 3, 2008, 8:06:56 PM, you wrote:
RE> Albert Chin wrote: >> On Thu, Jul 03, 2008 at 01:43:36PM +0300, Mertol Ozyoney wrote: >> >>> You are right that J series do not have nvram onboard. However most Jbods >>> like HPS's MSA series have some nvram. >>> The idea behind not using nvram on the Jbod's is >>> >>> -) There is no use to add limited ram to a JBOD as disks already have a lot >>> of cache. >>> -) It's easy to design a redundant Jbod without nvram. If you have nvram and >>> need redundancy you need to design more complex HW and more complex firmware >>> -) Bateries are the first thing to fail >>> -) Servers already have too much ram >>> >> >> Well, if the server attached to the J series is doing ZFS/NFS, >> performance will increase with zfs:zfs_nocacheflush=1. But, without >> battery-backed NVRAM, this really isn't "safe". So, for this usage case, >> unless the server has battery-backed NVRAM, I don't see how the J series >> is good for ZFS/NFS usage. >> >> RE> The zfs_nocacheflush problem should be mostly gone as the fix was RE> implemented in b74. We really expect that this recommendation will RE> disappear, except in its viral form. RE> http://bugs.opensolaris.org/view_bug.do?bug_id=6462690 Has anyone performed some study with well know disk arrays if they actually properly honor the bit? Unless the fix incorporates recognizing "bad" arrays and automatically is not sending cache flushes to their LUNs....? -- Best regards, Robert Milkowski mailto:[EMAIL PROTECTED] http://milek.blogspot.com _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
