Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
/ 2004-09-23 10:44:47 -0300 \ crsurf: > > Yes is it, when using database throughput not pass 6 Mb/s. so running your database on top of drbd gives 6 MB/sec throughput. resync has nothing to do with it. right? > > how do you measure? > > I?m using iptraf to view network traffic and is that wich says the > throughput I mencioned before (4~6 Mb when inserting in database and > 40~50 Mb/s when doing SyncingAll). > > This is the output of iostat on primary server > > avg-cpu: %user %nice %sys %iowait %idle > 0.79 0.00 1.54 3.28 94.39 > Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn > sda 120.44 5271.05 1872.72 290326040 103148100 > sda1 0.00 0.05 0.01 2928 424 > sda2 120.43 5271.00 1872.71 290323088 103147676 > > > This is the output of iostat on seconday server > > avg-cpu: %user %nice %sys %iowait %idle > 0.32 0.00 1.82 3.58 94.27 > Device: tps Blk_read/s Blk_wrtn/s Blk_read Blk_wrtn > sda 212.38 1159.70 6143.03 84933272 449898916 > sda1 0.00 0.02 0.01 1736 416 > sda2 212.38 1159.67 6143.02 84931456 449898500 this is almost completely irrelevant. I ask you to run your database benchmark on some local device, and on drbd, and compare the results. and using other well established benchmarks, like bonnie(++), tiobench ... might be helpful for you to find the actual bottleneck. again, run them on some local disk, and then on top of drbd, and compare the results. Lars Ellenberg -- please use the "List-Reply" function of your email client.