Disk array setup opinions
by Jerome, Ron
I am currently having a debate about the best way to carve up Dell MD3200's to be used as OST's in a Lustre file system and I invite this community to weigh in...
I am of the opinion that it should be setup as multiple raid groups each having a single LUN, with each raid group representing an OST, while my colleague feels that it should be setup as a single raid group across the whole array with multiple LUNS, with each LUN representing an OST.
Does anyone in this group have an opinion (one way or another)?
Regards,
Ron Jerome
9 years, 2 months
multi-rail setup - Lustre 2.1.4 - Whamcloud
by Edward Walter
Hello list,
We're attempting to setup a multi-rail lustre configuration so we can
provide access to the same lustre filesystem from two different
clusters. Both clusters are IB connected and we're using o2ib as the
protocol. The lustre servers have multiple IB cards each and are
connected to a separate IB switch on each cluster.
On cluster 1: we've got the following (simplified) setup:
> [root@mdt-3-40 ~]# lctl list_nids
> 172.16.1.113@o2ib
> 172.16.2.113@o2ib2
> [root@oss-0-19 ~]# lctl list_nids
> 172.16.1.103@o2ib
> 172.16.2.103@o2ib1
Clients from cluster 1 can ping these servers:
> [root@compute-1-5 data]# ping -c 1 172.16.1.113
> PING 172.16.1.113 (172.16.1.113) 56(84) bytes of data.
> 64 bytes from 172.16.1.113: icmp_seq=1 ttl=64 time=2.04 ms
>
> --- 172.16.1.113 ping statistics ---
> 1 packets transmitted, 1 received, 0% packet loss, time 0ms
> rtt min/avg/max/mdev = 2.049/2.049/2.049/0.000 ms
> [root@compute-1-5 data]# ping -c 1 172.16.1.103
> PING 172.16.1.103 (172.16.1.103) 56(84) bytes of data.
> 64 bytes from 172.16.1.103: icmp_seq=1 ttl=64 time=2.07 ms
>
> --- 172.16.1.103 ping statistics ---
> 1 packets transmitted, 1 received, 0% packet loss, time 0ms
> rtt min/avg/max/mdev = 2.079/2.079/2.079/0.000 ms
Clients from cluster 1 can also ping them using lnet:
> [root@compute-1-5 data]# lctl ping mdt-3-40.ib@o2ib0
> 12345-0@lo
> 12345-172.16.1.113@o2ib
> 12345-172.16.2.113@o2ib2
> [root@compute-1-5 data]# lctl ping oss-0-19.ib@o2ib0
> 12345-0@lo
> 12345-172.16.1.103@o2ib
> 12345-172.16.2.103@o2ib1
Finally clients from cluster1 can mount the lustre filesystem:
> [root@compute-1-5 data]# lfs check servers
> data-MDT0000-mdc-ffff81021df32800: active
> data-OST0000-osc-ffff81021df32800: active
On cluster 2: clients can ping the IPOIB addresses for these servers:
> [root@compute-1-1 ~]# ping -c 1 172.16.2.113
> PING 172.16.2.113 (172.16.2.113) 56(84) bytes of data.
> 64 bytes from 172.16.2.113: icmp_seq=1 ttl=64 time=0.096 ms
>
> --- 172.16.2.113 ping statistics ---
> 1 packets transmitted, 1 received, 0% packet loss, time 0ms
> rtt min/avg/max/mdev = 0.096/0.096/0.096/0.000 ms
> [root@compute-1-1 ~]# ping -c 1 172.16.2.103
> PING 172.16.2.103 (172.16.2.103) 56(84) bytes of data.
> 64 bytes from 172.16.2.103: icmp_seq=1 ttl=64 time=0.083 ms
>
> --- 172.16.2.103 ping statistics ---
> 1 packets transmitted, 1 received, 0% packet loss, time 0ms
> rtt min/avg/max/mdev = 0.083/0.083/0.083/0.000 ms
Doing an lctl ping fails though:
> [root@compute-1-1 ~]# lctl ping 172.16.2.103@o2ib
> failed to ping 172.16.2.103@o2ib: Input/output error
> [root@compute-1-1 ~]# lctl ping 172.16.2.113@o2ib
> failed to ping 172.16.2.113@o2ib: Input/output error
Pinging the client on cluster 2 (from itself) works though (so lnet is
up and working):
> [root@compute-1-1 ~]# lctl ping compute-1-1.ib@o2ib
> 12345-0@lo
> 12345-172.16.2.247@o2ib
The net result of all of this is that I'm getting messages like these
when I try to mount the lustre filesystem from cluster 2:
> LustreError: 1998:0:(o2iblnd_cb.c:2249:kiblnd_passive_connect()) Can't accept 172.16.2.247@o2ib on 172.16.1.113@o2ib (ib2:1:172.16.2.113): bad dst nid 172.16.2.113@o2ib
> LustreError: 1998:0:(o2iblnd_cb.c:2249:kiblnd_passive_connect()) Can't accept 172.16.2.247@o2ib on 172.16.1.113@o2ib (ib2:1:172.16.2.113): bad dst nid 172.16.2.113@o2ib
> LustreError: 1998:0:(o2iblnd_cb.c:2249:kiblnd_passive_connect()) Can't accept 172.16.2.247@o2ib on 172.16.1.113@o2ib (ib2:1:172.16.2.113): bad dst nid 172.16.2.113@o2ib
> LustreError: 1998:0:(o2iblnd_cb.c:2249:kiblnd_passive_connect()) Can't accept 172.16.2.247@o2ib on 172.16.1.113@o2ib (ib2:1:172.16.2.113): bad dst nid 172.16.2.113@o2ib
> LustreError: 1998:0:(o2iblnd_cb.c:2249:kiblnd_passive_connect()) Can't accept 172.16.2.247@o2ib on 172.16.1.113@o2ib (ib2:1:172.16.2.113): bad dst nid 172.16.2.113@o2ib
> LustreError: 1998:0:(o2iblnd_cb.c:2249:kiblnd_passive_connect()) Can't accept 172.16.2.247@o2ib on 172.16.1.113@o2ib (ib2:1:172.16.2.113): bad dst nid 172.16.2.113@o2ib
> LustreError: 1998:0:(o2iblnd_cb.c:2249:kiblnd_passive_connect()) Can't accept 172.16.2.247@o2ib on 172.16.1.113@o2ib (ib2:1:172.16.2.113): bad dst nid 172.16.2.113@o2ib
Here are the configs for the MDS server:
> echo 'options lnet ip2nets="o2ib0(ib0),o2ib2(ib2) 172.16.[1-2].*"' > /etc/modprobe.d/lustre.conf
> modprobe lnet
> mkfs.lustre --reformat --mgs /dev/MDT340/mgs
> mount -t lustre /dev/MDT340/mgs /lustre/mgs
> mkfs.lustre --reformat --mdt --fsname=data --mgsnode=mdt-3-40.ib@o2ib,mdt-3-40.coma-ib@o2ib2 /dev/MDT340/data
> mount -t lustre /dev/MDT340/data /lustre/data
These are the configs for the OSS server:
> echo 'options lnet ip2nets="o2ib0(ib0),o2ib1(ib1) 172.16.[1-2].*"' > /etc/modprobe.d/lustre.conf
> mkfs.lustre --reformat --fsname data --mgsnode=mdt-3-40.ib@o2ib,mdt-3-40.coma-ib@o2ib1 --ost /dev/sdb1
> mount -t lustre /dev/sdb1 /lustre/data-sdb1
Here's the config for the non-working client:
> [root@compute-1-1 ~]# cat /etc/modprobe.d/lustre.conf
> options lnet networks=o2ib0(ib0)
> [root@compute-1-1 ~]# mount -t lustre warp-mdt-3-40.ib@o2ib0:/data /root/data
> mount.lustre: mount warp-mdt-3-40.ib@o2ib0:/data at /root/data failed: Invalid argument
> This may have multiple causes.
> Is 'data' the correct filesystem name?
> Are the mount options correct?
> Check the syslog for more info.
Maybe I'm missing something obvious here. Any suggestions would be
appreciated.
Thanks much.
-Ed Walter
Carnegie Mellon University
9 years, 2 months
Lustre error
by Andersen, Per
We started getting huge numbers of these messages on one of our OSS servers, there are two OSS servers in the system each with 2 OSTs on them. The second OSS does not generate these messages.
Mar 5 08:20:20 rt-oss1 kernel: LustreError: 29297:0:(fsfilt-ldiskfs.c:493:fsfilt_ldiskfs_brw_start()) want too many journal credits (555) using 500 instead
Mar 5 08:20:20 rt-oss1 kernel: LustreError: 29297:0:(fsfilt-ldiskfs.c:493:fsfilt_ldiskfs_brw_start()) Skipped 16215 previous similar messages
Does anyone know what this message is about? Since we started getting them the lustre system performance has been reduced.
Per Andersen Ph.D. P.E.
HPCC
Texas Tech University
806-834-6924
9 years, 2 months
LUG 2012 Abstracts
by ssimms@iu.edu
Greetings!
We want you to keep those cards and letters coming. Because we haven't
heard from *you*, we're extending LUG Abstract submission until Friday
this week (March 8th, 2013).
For abstract submission, we are using the easychair conference system.
The website for LUG 2013 is here:
https://www.easychair.org/conferences/?conf=lug2013
We're really looking forward to seeing your contribution at this year's
meeting in San Diego!
Lustre User Group meeting (LUG)'13 will take place at the Omni Hotel in
the Downtown Gaslamp Quarter. LUG Sessions will be held from April 16-18.
Be sure to register by March 15, 2013, to take advantage of our early bird
registration discount rate of $525. Registration closes on Friday March
22, 2013. We have an interface this year that includes the full agenda,
hotel information and a lot more:
https://www.gaveledge.com/SFS1301/home
If you have questions or problems, feel free to mail Stephen Simms
(ssimms(a)iu.edu), LUG 2013 program chair.
Thanks so much for your attention!
Sincerely,
Stephen Simms
LUG 2013 Program Chair
ssimms(a)iu.edu
812-855-7211
9 years, 2 months
Lustre HSM git repo unavailable?
by Peng Tao
Hi,
I just noticed that there is a repo for lustre-hsm on
git.whamcloud.com. However, I cannot clone it (lustre-hsm.git) somehow
although I can access lustre-release.git w/o problems. So I'm
wondering if there is any git server issue or is the code unavailable
for download at the moment?
$git clone git://git.whamcloud.com/fs/lustre-hsm.git
Cloning into lustre-hsm...
fatal: The remote end hung up unexpectedly
--
Thanks,
Tao
9 years, 2 months