[DRBD-user] suffering from iowait

Mark Watts m.watts at eris.qinetiq.com
Fri Aug 15 09:51:30 CEST 2008

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


On Thursday 14 August 2008 23:04:47 Wolfram Schlich wrote:
> Hi,
>
> imagine the following setup:
> - 2 node cluster
> - 2x Dell PowerEdge 860, Dual Core Xeon 2.13GHz,
>   4GB RAM, SATA HW RAID1 (write cache enabled,
>   NCQ enabled), 1GE link for internal cluster
>   communication (DRBD, Heartbeat etc.), Heartbeat
>   fencing using IPMI-over-LAN
> - Linux 2.6.25 x86_64 with deadline or CFQ scheduler
>   (tried both)
> - DRBD 8.2.6
> - Heartbeat 2.1.3 with dopd enabled
> - drbd.conf: see http://paste.frubar.net/8786
> - MySQL-5.0.60 with sync-binlog=1 and
>   innodb_flush_log_at_trx_commit=1
>
> With the DRBD instances connected, the system
> spends >30% on iowait and is >30% idle
> during a load test.
>
> With the secondary instance disconnected, the system
> does (almost) not spend any time on iowait, CPU
> usage is at around 80% user and 20% system, but
> the load goes slowly up to more than 20 after a
> minute or so...
>
> What could be the reason for both issues (iowait
> vs. load skyrocketing)?
>
> Thanks!

If its like other Dell's, you should start by disabling the TCP Offload (TOE) 
by removing the hardware key on the motherboard.

Mark.

-- 
Mark Watts BSc RHCE MBCS
Senior Systems Engineer
QinetiQ Applied Technologies
GPG Key: http://www.linux-corner.info/mwatts.gpg
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20080815/27d986ad/attachment.pgp>


More information about the drbd-user mailing list