Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Lars Ellenberg wrote: > > On Fri, Feb 11, 2011 at 02:16:06AM -0800, ionral wrote: >> >> >Do not use "no-disk-drain", unless you are absolutely sure that nothing >> >in the IO stack below DRBD could possibly reorder writes. >> >> in my case I really think so or not?? >> >> I have this situation: >> >> sda -> drbd0 -> pv -> vg -> lv (virtual disks for virtual machine) > > actually it is sda -> linux block layer elevator -> drbd, > so there is reordering involved there. > > and depending on what exactly sda is (some external raid box?) > there may very well be reordering there, too. > > So, no. Just don't. > ok, I did not know ... then no-disk-drain should never be used ... your advice then is to use anything but this parameter right? or rather, I usually 3 types of hardware: 1. two nodes with hardware raid controller with BBU 2. two nodes with hardware raid controller without BBU 3. two nodes without hardware raid controller and implementations of software raid (raid0) in these two cases, it is the best configuration for performance? (without allowing dangerous situations of data loss of course) thanks -- View this message in context: http://old.nabble.com/DRBD-write-after-write-%28cache%29-tp30850998p30902134.html Sent from the DRBD - User mailing list archive at Nabble.com.