<html><head><style type='text/css'>p { margin: 0; }</style></head><body><div style='font-family: Arial; font-size: 12pt; color: #000000'>Hello Sebastian<div><br></div><div>Thanks for the hint. The be sure that the settings are exactly the same on both sides, I wrote a script to set these settings by ssh on both nodes equally. So the current value for max_sectors_kb is: 64 on both nodes. But I played around with these values and nothing helped. When the nodes are connected, the write speed is even lower than in single node mode. I think there must be some kind of bottleneck between the blocksizes / ioscheduler, etc. and drbd but could find the igniting hint. </div><div><br></div><div>Strange is also that the sync speed is very low (triggered manually resync to retest that) 66MB /s. All tests with different config settings showed no performance changement. blktrace also shows nothing more flashy which could lead to more. while resyncing and dd'ing the iostat shows 1.07% iowat which is imho very low value for a system under full io load. </div><div><br></div><div>Is there a possibility to debug the network usage and internal network stats of drbd besides drbdsetup show 0?</div><div><br></div><div>Regards</div><div><br></div><div>Roland</div><div><br></div><div><br></div><div><br><hr id="zwchr"><b>Von: </b>"Sebastian Riemer" <sebastian.riemer@profitbricks.com><br><b>An: </b>"Roland Kaeser" <roland.kaeser@ziil.ch><br><b>CC: </b>drbd-user@lists.linbit.com<br><b>Gesendet: </b>Mittwoch, 26. September 2012 15:49:07<br><b>Betreff: </b>Re: [DRBD-user] Write performance again<br><br>Hi Roland,<br><br>I've analysed the DRBD 8.4.1 kernel code some time ago.<br><br>DRBD has a dynamic IO request size detection which always starts at 4<br>KiB. If you've got a connection to the other system, then it negotiates<br>a higher IO request size (128 KiB because of hashing functions). If you<br>loose the connection, it's 4 KiB again.<br><br>You can use the tool "blktrace" to trace how big the IO requests are. Or<br>you can look at /sys/block/drbd0/queue/max_sectors_kb.<br><br>IMHO this must become settable to the limits from the devices below if<br>you already know that "max_sectors_kb" is completely equal on both sides.<br><br>See:<br>http://www.gossamer-threads.com/lists/drbd/users/24104<br><br>Cheers,<br>Sebastian<br><br><br><br>On 26.09.2012 14:50, Roland Kaeser wrote:<br>> Only the drbd is very slow. (Single node, just to test write<br>> perfomance on local block device.<br><br><span><br><br>-- <br><span name="x"></span><div><meta http-equiv="content-type" content="text/html; charset=utf-8"><font size="2"><a href="http://www.ziil.ch">Ziil Informatiklösungen</a> GmbH</font><br><font size="2"><span style="font-weight: bold;">Roland Käser</span></font></div><div><font size="2"><span style="font-weight: bold;"></span>Leiter Technik<br>Romanshornerstr. 134, <br>CH-8280 Kreuzlingen</font><br><font size="2">Tel: +41 71 671 27 78</font></div><div><font size="2">Fax: +41 71 671 27 79<br>www.ziil.ch</font></div><span name="x"></span><br></span></div></div></body></html>