<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; charset=iso-8859-1">
<META NAME="Generator" CONTENT="MS Exchange Server version 6.5.7653.38">
<TITLE>RE: [DRBD-user] drbd 0.7 vs 8 latency</TITLE>
</HEAD>
<BODY>
<!-- Converted from text/plain format -->
<P><FONT SIZE=2>On another note, what would probably be the best ones to use on a non-battery backup, SATA RAID setup for DRBD? And will adding these to an existing configuration cause any issues?<BR>
<BR>
Mike Sweetser<BR>
<BR>
-----Original Message-----<BR>
From: drbd-user-bounces@lists.linbit.com on behalf of X LAci<BR>
Sent: Thu 1/1/2009 12:14 PM<BR>
To: drbd-user@lists.linbit.com<BR>
Subject: Re: [DRBD-user] drbd 0.7 vs 8 latency<BR>
<BR>
Dear all,<BR>
<BR>
Happy new year to all who is celebrating.<BR>
<BR>
Lars was right.<BR>
<BR>
no-disk-barrier;<BR>
no-disk-flushes;<BR>
no-md-flushes;<BR>
<BR>
All three of these was needed to bring back the io performance to the 0.7 level.<BR>
I've read in the documentation that these should be used only on BBWC RAID,<BR>
but I didn't know that a simple SATA disk based system also needs this for higher performance.<BR>
<BR>
By default 8.3.0 uses disk-barrier, if one disables this, then it uses flush, if that is<BR>
disabled also, then comes the drain method.<BR>
<BR>
Thanks for the help.<BR>
<BR>
Best regards,<BR>
<BR>
lanlaf<BR>
<BR>
<BR>
On Wed, 2008-12-31 at 16:18 +0100, Lars Ellenberg wrote:<BR>
> On Wed, Dec 31, 2008 at 12:52:41PM +0100, X LAci wrote:<BR>
> > Dear all,<BR>
> ><BR>
> > I ran into high latency with drbd 8.3.0, on the same hardware where drbd<BR>
> > 0.7 was OK.<BR>
> ><BR>
> > I have created a test system Yesterday, two Lenovo Thinkcenter PCs, P4<BR>
> > 3.0 GHz, Intel Chipset, 1 SATA disk in each PC, Gigabit Ethernet, one<BR>
> > cable between them. I've installed Debian Etch on each of them, and drbd<BR>
> > 0.7 that came with Etch. Drbd 8.3.0 was compiled by me.<BR>
> ><BR>
> > One goal was to verify that 0.7 -> 8 version upgrade goes without data<BR>
> > loss. That is succeeded, there was no data loss.<BR>
> ><BR>
> > I have tested the performance as described in the performance tuning<BR>
> > webinar.<BR>
> ><BR>
> > Hdparm and dd showed that the disks are capable of reading and writing<BR>
> > at about 62 MByte/sec.<BR>
> ><BR>
> > The disk latency for each node:<BR>
> ><BR>
> > dd if=/dev/zero of=/dev/sda3 bs=512 count=1000 oflag=direct<BR>
> ><BR>
> > node1: 512000 bytes (512 kB) copied, 0.158945 seconds, 3.2 MB/s<BR>
> > node2: 512000 bytes (512 kB) copied, 0.158367 seconds, 3.2 MB/s<BR>
> ><BR>
> > For initial sync, I put resync rate to 60 MByte/sec so that it finishes<BR>
> > quickly. The actual resync rate was around 48 MByte/sec.<BR>
> ><BR>
> > Throughput was OK with 8.3.0, bs=300M count=1, 60.3 MByte/sec.<BR>
> ><BR>
> > Latency with drbd 0.7 was acceptable also:<BR>
> ><BR>
> > dd if=/dev/zero of=/dev/drbd0 bs=512 count=1000 oflag=direct<BR>
> > 512000 bytes (512 kB) copied, 0.308893 seconds, 1.7 MB/s<BR>
> ><BR>
> > Latency with drbd 8.3.0 was very bad:<BR>
> ><BR>
> > dd if=/dev/zero of=/dev/drbd0 bs=512 count=1000 oflag=direct<BR>
> > 512000 bytes (512 kB) copied, 8.36032 seconds, 61.2 kB/s<BR>
> ><BR>
> > It was not resyncing, there was no other activity on the PCs.<BR>
> ><BR>
> > I've also tried with drbd 8.2.6, same results. Also tried with Debian<BR>
> > Etch 2.6.18 and 2.6.24-etchnhalf kernel, same results.<BR>
> ><BR>
> > drbd.conf was basicly default, only put in disk and network parameters,<BR>
> > resync rate was changed to 20 Mbyte/sec, and al-extents 1201.<BR>
> ><BR>
> > What could cause this problem?<BR>
><BR>
> drbd 0.7 has no idea about disk flushes or barriers,<BR>
> so on volatile caches it may cause data loss on power outage.<BR>
><BR>
> drbd 8 by default cares very much about explicitly flushing<BR>
> or inserting barriers into the data stream to avoid this problem.<BR>
> however when running on huge (controller) caches, and if the<BR>
> implementation of "flush" is indeed a full cache flush,<BR>
> these frequent flushes can degrade performance considerably.<BR>
><BR>
> they are not even necessary when your cache is batter backed.<BR>
><BR>
> so if you are running on a "safe" device<BR>
> (either NON-volatile, battery backed, cache; or no cache at all),<BR>
> consider turning these additional flushes off:<BR>
><BR>
> no-disk-barrier;<BR>
> no-disk-flushes;<BR>
> no-md-flushes;<BR>
><BR>
><BR>
><BR>
<BR>
_______________________________________________<BR>
drbd-user mailing list<BR>
drbd-user@lists.linbit.com<BR>
<A HREF="http://lists.linbit.com/mailman/listinfo/drbd-user">http://lists.linbit.com/mailman/listinfo/drbd-user</A><BR>
<BR>
<BR>
</FONT>
</P>
</BODY>
</HTML>