<table width=100% dir="ltr"><tr><td><DIV id=wrteplaceholder name="wrteplaceholder"> </DIV>
<DIV name="wrteplaceholder">We had the same problem on our system. We used to have one partition that held all the data we need to share between the primary and standby servers.</DIV>
<DIV name="wrteplaceholder">The data could be roughly divided into server related data and NFS exported data. The problem we encountered is an overflow of the NFS exported data</DIV>
<DIV name="wrteplaceholder">prevented the server from functioning because it lacked space for the local data.</DIV>
<DIV name="wrteplaceholder">Consequently, we are now moving to a 2-partition configuration. Drbd0 is used for the local data and drbd1 for the NFS exported data.</DIV>
<DIV name="wrteplaceholder"> </DIV>
<DIV name="wrteplaceholder">Regarding your concern about sync time, we have upgraded the system to 0.6.11 which contains the sync-group directive in drbd.conf. It allows you to</DIV>
<DIV name="wrteplaceholder">have an order in the syncing activity, to prevent intereference between syncing activities of the partitions. It syncs drbd0 first and then drbd1.</DIV>
<DIV name="wrteplaceholder">The sync time is no different than with one partition, since the overall amount of data you need to synchronize is the same. We are now putting the new configuration into production, so I hope to have results soon as to whether the new configuration proves itself in the field. Testing showed that it works well.</DIV>
<DIV name="wrteplaceholder"><BR>Relating to your configuration, our disk is smaller. We are also using ext3 file systems..<BR></DIV>
<DIV name="wrteplaceholder">Doron Zifrony</DIV>
<DIV name="wrteplaceholder">BladeFusion</DIV>
<DIV name="wrteplaceholder"><A href="mailto:dhz_bf@walla.co.il">dhz_bf@walla.co.il</A></DIV>
<DIV name="wrteplaceholder"> </DIV>
<DIV name="wrteplaceholder">
<TABLE style="BORDER-RIGHT: 1px outset; BORDER-TOP: 1px outset; BORDER-LEFT: 1px outset; BORDER-BOTTOM: 1px outset" cellSpacing=0 cellPadding=0 width="100%" bgColor=#f0f0f0>
<TBODY>
<TR>
<TD><FONT size=2><B><hmalte@you.oih.rwth-aachen.de></B><BR>[DRBD-user] disk/partition/resource layout</FONT></TD></TR></TBODY></TABLE></DIV>
<DIV name="wrteplaceholder"><PRE dir=ltr><BR>I've got two servers (x-connected via gigabit-NICs) which
<BR>should provide high availabilty NFS-service for a number of
<BR>web/mail/something front end servers.
<BR>
<BR>I am currently planning on how to layout the existing
<BR>disk space:
<BR>
<BR>a.) Multiple partitions "connected" via multiple
<BR>drbd-resources (drbd0, drbd1, ...). NFS-Exports:
<BR>/mnt/drbd0/ (mail), /mnt/drbd1/ (www), ...
<BR>
<BR>b.) One big partition on one resource (drbd0). NFS-
<BR>Exports: /mnt/drbd0/mail/, /mnt/drbd0/www/, ...
<BR>
<BR>Possible (Dis-)Advantages:
<BR>
<BR>a.)
<BR>Pro: Potential file system corruption only affects one
<BR>partition and corresponding front end servers.
<BR>Con: Not recommended in DRBD documentation. First tests
<BR>show poor sync(-all) performance (as exptected).
<BR>
<BR>b.)
<BR>Pro: Easy "DRBD-doc-recommended" setup.
<BR>Con: Potential file system corruptions cause failures
<BR>on all frontend servers.
<BR>
<BR>
<BR>Total disk size on each server is ~200GB. Expected number
<BR>of files: >800.000 with moderate I/O. Preferred file system
<BR>is ext3.
<BR>
<BR>Any suggestions are welcome.
<BR>
<BR>H.
<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></PRE></DIV>
<DIV name="wrteplaceholder">
<HR noShade SIZE=1>
</DIV>
<DIV name="wrteplaceholder"><BR><BR></DIV>
<DIV name="wrteplaceholder">
<HR SIZE=1>
</DIV></td></tr></table>
<br>
<hr size=1 noshade>
Walla! Mail, Get Your Private, Free E-mail from Walla! at:<br>
<a href="http://mail.walla.co.il">http://mail.walla.co.il</a><br>