<HTML>
<HEAD>
<TITLE>Re: [Lustre-discuss] More failover issues</TITLE>
</HEAD>
<BODY>
<FONT SIZE="5"><FONT FACE="Calibri, Verdana, Helvetica, Arial"><SPAN STYLE='font-size:9pt'>Yes only one MGS per site, but you should be able to specify multiple MGS nodes. We have done it before with 1.6.0. See <a href="http://manual.lustre.org/manual/LustreManual16_HTML/DynamicHTML-05-1.html">http://manual.lustre.org/manual/LustreManual16_HTML/DynamicHTML-05-1.html</a> section 2.2.2.1.<BR>
<BR>
Robert<BR>
<BR>
<BR>
On 11/12/07 2:48 PM, "Wojciech Turek" <wjt27@cam.ac.uk> wrote:<BR>
<BR>
</SPAN></FONT></FONT><BLOCKQUOTE><FONT SIZE="5"><FONT FACE="Calibri, Verdana, Helvetica, Arial"><SPAN STYLE='font-size:9pt'>Hi,<BR>
<BR>
I think this is because there can be only one MGS per lustre installation (this is what manual says).<BR>
<BR>
Wojciech Turek<BR>
On 12 Nov 2007, at 21:18, Robert LeBlanc wrote:<BR>
<BR>
</SPAN></FONT></FONT><BLOCKQUOTE><FONT SIZE="5"><FONT FACE="Calibri, Verdana, Helvetica, Arial"><SPAN STYLE='font-size:9pt'> <BR>
<BR>
This is what I'm getting:<BR>
<BR>
head2-2:~# mkfs.lustre --mkfsoptions="-O dir_index" --reformat --mdt --fsname=home --mgsnode=192.168.1.252@o2ib --mgsnode=192.168.1.253@o2ib --failnode=192.168.1.252@o2ib /dev/mapper/ldiskd-part1<BR>
<BR>
Permanent disk data:<BR>
Target: home-MDTffff<BR>
Index: unassigned<BR>
Lustre FS: home<BR>
Mount type: ldiskfs<BR>
Flags: 0x71<BR>
(MDT needs_index first_time update )<BR>
Persistent mount opts: errors=remount-ro,iopen_nopriv,user_xattr<BR>
Parameters: mgsnode=192.168.1.253@o2ib failover.node=192.168.1.252@o2ib mdt.group_upcall=/usr/sbin/l_getgroups<BR>
<BR>
device size = 972MB<BR>
formatting backing filesystem ldiskfs on /dev/mapper/ldiskd-part1<BR>
target name home-MDTffff<BR>
4k blocks 0<BR>
options -O dir_index -i 4096 -I 512 -q -F<BR>
mkfs_cmd = mkfs.ext2 -j -b 4096 -L home-MDTffff -O dir_index -i 4096 -I 512 -q -F /dev/mapper/ldiskd-part1<BR>
Writing CONFIGS/mountdata<BR>
<BR>
<BR>
For some reason, only the last --mgsnode option is being kept.<BR>
<BR>
Robert<BR>
<BR>
<BR>
-----Original Message-----<BR>
From: Nathan Rutman [<a href="mailto:Nathan.Rutman@Sun.COM]">mailto:Nathan.Rutman@Sun.COM]</a><BR>
Sent: Mon 11/12/2007 1:51 PM<BR>
To: Robert LeBlanc<BR>
Cc: lustre<BR>
Subject: Re: [Lustre-discuss] More failover issues<BR>
<BR>
Robert LeBlanc wrote:<BR>
> In 1.6.0, when creating a MDT, you could specify multiple --mgsnode options<BR>
> and it would failover between them. 1.6.3 only seems to take the last one<BR>
> and --mgsnode=192.168.1.252@o2ib:192.168.1.253@o2ib doesn't seem to failover<BR>
> to the other node. Any ideas how to get around this?<BR>
><BR>
Multiple --mgsnode parameters should work:<BR>
mkfs.lustre --mkfsoptions="-O dir_index" --reformat --mdt<BR>
--mgsnode=192.168.1.253@o2ib --mgsnode=1@elan --device-size=10000 /tmp/foo<BR>
<BR>
Permanent disk data:<BR>
Target: lustre-MDTffff<BR>
Index: unassigned<BR>
Lustre FS: lustre<BR>
Mount type: ldiskfs<BR>
Flags: 0x71<BR>
(MDT needs_index first_time update )<BR>
Persistent mount opts: errors=remount-ro,iopen_nopriv,user_xattr<BR>
Parameters: mgsnode=192.168.1.253@o2ib mgsnode=1@elan<BR>
<BR>
> Robert<BR>
><BR>
> Robert LeBlanc<BR>
> College of Life Sciences Computer Support<BR>
> Brigham Young University<BR>
> leblanc@byu.edu<BR>
> (801)422-1882<BR>
><BR>
><BR>
> _______________________________________________<BR>
> Lustre-discuss mailing list<BR>
> Lustre-discuss@clusterfs.com<BR>
> https://mail.clusterfs.com/mailman/listinfo/lustre-discuss<BR>
><BR>
<BR>
<BR>
<BR>
<BR>
_______________________________________________<BR>
Lustre-discuss mailing list<BR>
Lustre-discuss@clusterfs.com<BR>
https://mail.clusterfs.com/mailman/listinfo/lustre-discuss<BR>
<BR>
<BR>
<BR>
</SPAN></FONT></FONT><FONT SIZE="6"><FONT FACE="Lucida Grande"><SPAN STYLE='font-size:10pt'>Mr Wojciech Turek<BR>
</SPAN></FONT></FONT><FONT SIZE="5"><FONT FACE="Helvetica, Verdana, Arial"><SPAN STYLE='font-size:9pt'>Assistant System Manager<BR>
</SPAN></FONT></FONT><FONT SIZE="6"><FONT FACE="Lucida Grande"><SPAN STYLE='font-size:10pt'>University of Cambridge<BR>
High Performance Computing service<BR>
</SPAN></FONT></FONT><FONT FACE="Lucida Grande"><FONT SIZE="5"><SPAN STYLE='font-size:8.5pt'>email: wjt27@cam.ac.uk<BR>
</SPAN></FONT><FONT SIZE="6"><SPAN STYLE='font-size:10pt'>tel. +441223763517<BR>
</SPAN></FONT></FONT><FONT SIZE="5"><FONT FACE="Helvetica, Verdana, Arial"><SPAN STYLE='font-size:9pt'><BR>
<BR>
<BR>
</SPAN></FONT><SPAN STYLE='font-size:9pt'><FONT FACE="Calibri, Verdana, Helvetica, Arial"><BR>
<BR>
</FONT></SPAN></FONT></BLOCKQUOTE></BLOCKQUOTE><FONT SIZE="5"><SPAN STYLE='font-size:9pt'><FONT FACE="Calibri, Verdana, Helvetica, Arial"><BR>
<BR>
Robert LeBlanc<BR>
College of Life Sciences Computer Support<BR>
Brigham Young University<BR>
leblanc@byu.edu<BR>
(801)422-1882<BR>
<BR>
</FONT></SPAN></FONT>
</BODY>
</HTML>