Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Mixter wrote: > > Well, I'm happy to report I think I solved the oopses, thanks to > Phillip Reisner's clue. > Now just still lost with the SLOW SLOW initial synchronisation between devices. <SNIP> > But with regards to the SLOW initializing/synching problem, see below, > I still have > no clue or fix. A benchmark was attached previously, dma is on and > all, I'm happy > for any clues on known drbd issues or (non-lethal!) hdparm commands... > < Prepare for Philipp to correct me ... but > So from some of the data below can I assume you have both DRBD devices on the same spindle, i.e., hda? Under 0.6.X (yes that is what I am still using) you could define each device to be in a different sync group (sync-group = [01]) in the drbd.conf. Even though my 7 DRBD devices were on a 5 disk raid5 set, putting each one into it's own sync group more than doubled the through put on sync. I have not read any of the code for 0.7.X but I suspect that trying to sync both at the same time means you have 4 seeks going on, 1) read data from DRBD 0 and send to slave 2) read data from DRBD 1 and send to slave 3) mark DRBD 0 internal log (ALs?) indicating section synced 4) mark DRBD 1 internal log (ALs?) indicating section synced And all this is happening on one spindle, with one set of heads. Philipp, Is there a chance you guys will ever setup 0.7.X so that the ALs can be a file in an alternate filesystem? Granted I will probably repartition the machines boot drives before advancing to 0.7.X, but it would make adding new DRBDs after initial setup easier if the AL could be placed somewhere with out having to repartition or use LVM. (Oh, yes I was meaning a non pay change... If I really needed it badly I suppose I could send LINBIT some money :) > version: 0.7.6 (api:77/proto:74) > SVN Revision: 1663 build by root at primary, 2005-01-24 11:08:48 > 0: cs:SyncTarget st:Secondary/Secondary ld:Inconsistent > ns:0 nr:15936 dw:15936 dr:0 al:0 bm:130 lo:0 pe:0 ua:0 ap:0 > [>...................] sync'ed: 0.1% (122282/122298)M > finish: 108:41:45 speed: 284 (240) K/sec > 1: cs:SyncTarget st:Secondary/Secondary ld:Inconsistent > ns:0 nr:15720 dw:15720 dr:0 al:0 bm:128 lo:0 pe:0 ua:0 ap:0 > [>...................] sync'ed: 0.1% (120009/120024)M > finish: 106:40:30 speed: 244 (240) K/sec > <SNIP> > /dev/hda: > Timing buffered disk reads: 8 MB in 3.07 seconds = 2.60 MB/sec > > /dev/hda: > Timing buffer-cache reads: 492 MB in 2.01 seconds = 245.18 MB/sec -- Todd Denniston Crane Division, Naval Surface Warfare Center (NSWC Crane) Harnessing the Power of Technology for the Warfighter