Hi,
Is it possible with Lustre to configure MDS/MDT/OST on the same node ?
I would appreciate an example of creating OST on MDS please.
I am trying to create OST by following the manual but I get into
following problem.
[root@storage01 ~]# mkfs.lustre --fsname datafs --ost0 --
mgsnode=mds16@tcp0 /dev/sdb
mkfs.lustre: unrecognized option `--ost0''
mkfs.lustre v1.6.0.1
usage: mkfs.lustre <target types> [options] <device>
<device>:block device or file (e.g /dev/sda or /tmp/ost1)
target types:
--ost: object storage, mutually exclusive with mdt,mgs
--mdt: metadata storage, mutually exclusive with ost
--mgs: configuration management service - one per site
options (in order of popularity):
--mgsnode=<nid>[,<...>] : NID(s) of a remote mgs
node
required for all targets other than the mgs
node
--fsname=<filesystem_name> : default is
''lustre''
--failnode=<nid>[,<...>] : NID(s) of a failover
partner
--param <key>=<value> : set a permanent parameter
e.g. --param sys.timeout=40
--param lov.stripesize=2M
--index=#N : target index (i.e. ost index within the
lov)
--comment=<user comment>: arbitrary user string (824
bytes)
--mountfsoptions=<opts> : permanent mount options
--backfstype=<fstype> : backing fs type (ext3, ldiskfs)
--device-size=#N(KB) : device size for loop devices
--mkfsoptions=<opts> : format options
--reformat: overwrite an existing disk
--stripe-count-hint=#N : used for optimizing MDT
inode size
--dryrun: just report what we would do; don''t write
to disk
--verbose : e.g. show mkfs progress
--quiet
mkfs.lustre: exiting with 1 (Operation not permitted)
Thanks for help
Wojciech Turek
On Wed, 2007-07-11 at 20:16 +0100, Wojciech Turek wrote:> Hi, > > Is it possible with Lustre to configure MDS/MDT/OST on the same node ?Just to be clear about nomenclature, and MDT (target) is a block device on an MDS (server) that lustre uses for metadata. MDSs hosts MDTs and OSSes host OSTs.> [root@storage01 ~]# mkfs.lustre --fsname datafs --ost0 -- > mgsnode=mds16@tcp0 /dev/sdb > mkfs.lustre: unrecognized option `--ost0''This error message says it all. --ost0 is not a valid option. The target types listed in the usage message says to use "--ost". Note the lack of the zero on the end of that. b.
Hi, On 11 Jul 2007, at 20:21, Brian J. Murrell wrote:> On Wed, 2007-07-11 at 20:16 +0100, Wojciech Turek wrote: >> Hi, >> >> Is it possible with Lustre to configure MDS/MDT/OST on the same >> node ? > > Just to be clear about nomenclature, and MDT (target) is a block > device > on an MDS (server) that lustre uses for metadata. MDSs hosts MDTs and > OSSes host OSTs.My mistake I''ve made typo I actually mean MGS/MDT/and OST on the same node> >> [root@storage01 ~]# mkfs.lustre --fsname datafs --ost0 -- >> mgsnode=mds16@tcp0 /dev/sdb >> mkfs.lustre: unrecognized option `--ost0'' > > This error message says it all. --ost0 is not a valid option. The > target types listed in the usage message says to use "--ost". Note > the > lack of the zero on the end of that.Ok but that was pasted from Lustre 1.6 manual. On the node I created MGS and MDT on /dev/sda6, also on the same node I have four other block devices sdb,sdc,sdd,sde, how I should construct the command line to create four OST for this block devices.> > b. > > > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss@clusterfs.com > https://mail.clusterfs.com/mailman/listinfo/lustre-discussThanks Wojciech Turek
On Wed, 2007-07-11 at 20:45 +0100, Wojciech Turek wrote:> Ok but that was pasted from Lustre 1.6 manual.Oh, this is not good then. I will look into that.> On the node I created MGS and MDT on /dev/sda6,OK.> also on the same node > I have four other block devices sdb,sdc,sdd,sde, how I should > construct the command line to create four OST for this block devices.Same command line for all four, just substitute the correct devices. Then create a mountpoint for them all "mkdir /mnt/ost{0,1,2,3}" and then mount the devices under each OST mount point. b.
Hi, I have a following command line: mkfs.lustre --fsname datafs --ost --mgsnode=storage01@tcp0 /dev/sdb Does --mgsnode above is correctly specified if MGS resides on the same node that OST? Host-name of the node is storage01 and modprobe.conf contains following line: options lnet networks=tcp0(eth0) cheers Wojciech Turek On 11 Jul 2007, at 20:54, Brian J. Murrell wrote:> On Wed, 2007-07-11 at 20:45 +0100, Wojciech Turek wrote: >> Ok but that was pasted from Lustre 1.6 manual. > > Oh, this is not good then. I will look into that. > >> On the node I created MGS and MDT on /dev/sda6, > > OK. > >> also on the same node >> I have four other block devices sdb,sdc,sdd,sde, how I should >> construct the command line to create four OST for this block devices. > > Same command line for all four, just substitute the correct devices. > Then create a mountpoint for them all "mkdir /mnt/ost{0,1,2,3}" and > then > mount the devices under each OST mount point. > > b. > > > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss@clusterfs.com > https://mail.clusterfs.com/mailman/listinfo/lustre-discuss
On Wed, 2007-07-11 at 21:07 +0100, Wojciech Turek wrote:> Hi, > > I have a following command line: > > mkfs.lustre --fsname datafs --ost --mgsnode=storage01@tcp0 /dev/sdb > > Does --mgsnode above is correctly specified if MGS resides on the > same node that OST? > Host-name of the node is storage01 and modprobe.conf contains > following line: options lnet networks=tcp0(eth0)That looks correct. b.
Hi, I mount MDT on the node [root@storage01 ~]# mount -t lustre /dev/sda6 /mnt/data/mdt then all OST mount -t lustre /dev/sdb /mnt/data/ost0 mount -t lustre /dev/sdc /mnt/data/ost1 mount -t lustre /dev/sdd /mnt/data/ost2 mount -t lustre /dev/sde /mnt/data/ost3 [root@storage01 ~]# lctl dl 0 UP mgs MGS MGS 5 1 UP mgc MGC10.143.245.1@tcp 7e101d66-fa88-bd81-896f-1601494e308e 5 2 UP mdt MDS MDS_uuid 3 3 UP lov datafs-mdtlov datafs-mdtlov_UUID 4 4 UP mds datafs-MDT0000 datafs-MDT0000_UUID 5 5 UP osc datafs-OST0000-osc datafs-mdtlov_UUID 5 6 UP osc datafs-OST0001-osc datafs-mdtlov_UUID 5 7 UP osc datafs-OST0002-osc datafs-mdtlov_UUID 5 8 UP osc datafs-OST0003-osc datafs-mdtlov_UUID 5 9 UP ost OSS OSS_uuid 3 10 UP obdfilter datafs-OST0000 datafs-OST0000_UUID 5 11 UP obdfilter datafs-OST0001 datafs-OST0001_UUID 5 12 UP obdfilter datafs-OST0002 datafs-OST0002_UUID 3 13 UP obdfilter datafs-OST0003 datafs-OST0003_UUID 3 [root@storage01 ~]# Then when I try to mount it on client node I get following problem: [root@node-a01 ~]# mount -t lustre storage01@tcp0:/datafs /mnt/datafs mount.lustre: mount 10.143.245.1@tcp:/datafs at /mnt/datafs failed: Input/output error Is the MGS running? the IP address is OK and it is pingable from the client node. [root@storage01 ~]# ping storage01 PING storage01.storage.cluster (10.143.245.1) 56(84) bytes of data. 64 bytes from storage01.storage.cluster (10.143.245.1): icmp_seq=0 ttl=64 time=0.022 ms 64 bytes from storage01.storage.cluster (10.143.245.1): icmp_seq=1 ttl=64 time=0.014 ms 64 bytes from storage01.storage.cluster (10.143.245.1): icmp_seq=2 ttl=64 time=0.013 ms Client node has LNET entry in modprobe.cfg for the right eth interface. options lnet networks=tcp0(eth0) Why it isn''t working? cheers Wojciech Turek On 11 Jul 2007, at 21:15, Brian J. Murrell wrote:> On Wed, 2007-07-11 at 21:07 +0100, Wojciech Turek wrote: >> Hi, >> >> I have a following command line: >> >> mkfs.lustre --fsname datafs --ost --mgsnode=storage01@tcp0 /dev/sdb >> >> Does --mgsnode above is correctly specified if MGS resides on the >> same node that OST? >> Host-name of the node is storage01 and modprobe.conf contains >> following line: options lnet networks=tcp0(eth0) > > That looks correct. > > b. > > > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss@clusterfs.com > https://mail.clusterfs.com/mailman/listinfo/lustre-discuss
On Thu, 2007-07-12 at 01:18 +0100, Wojciech Turek wrote:> Why it isn''t working?Let''s find out. First, from the client can you do: # lctl ping storage01 and tell me what the output is. Also, what does your client node''s syslog (/var/log/messages perhaps) say at the time of the mount? b.
Hi, I''ve found the problem. It is firewall on MDS cheers Wojciech Turek On 12 Jul 2007, at 01:18, Wojciech Turek wrote:> Hi, > > I mount MDT on the node > [root@storage01 ~]# mount -t lustre /dev/sda6 /mnt/data/mdt > then all OST > mount -t lustre /dev/sdb /mnt/data/ost0 > mount -t lustre /dev/sdc /mnt/data/ost1 > mount -t lustre /dev/sdd /mnt/data/ost2 > mount -t lustre /dev/sde /mnt/data/ost3 > > [root@storage01 ~]# lctl dl > 0 UP mgs MGS MGS 5 > 1 UP mgc MGC10.143.245.1@tcp 7e101d66-fa88-bd81-896f-1601494e308e 5 > 2 UP mdt MDS MDS_uuid 3 > 3 UP lov datafs-mdtlov datafs-mdtlov_UUID 4 > 4 UP mds datafs-MDT0000 datafs-MDT0000_UUID 5 > 5 UP osc datafs-OST0000-osc datafs-mdtlov_UUID 5 > 6 UP osc datafs-OST0001-osc datafs-mdtlov_UUID 5 > 7 UP osc datafs-OST0002-osc datafs-mdtlov_UUID 5 > 8 UP osc datafs-OST0003-osc datafs-mdtlov_UUID 5 > 9 UP ost OSS OSS_uuid 3 > 10 UP obdfilter datafs-OST0000 datafs-OST0000_UUID 5 > 11 UP obdfilter datafs-OST0001 datafs-OST0001_UUID 5 > 12 UP obdfilter datafs-OST0002 datafs-OST0002_UUID 3 > 13 UP obdfilter datafs-OST0003 datafs-OST0003_UUID 3 > [root@storage01 ~]# > > Then when I try to mount it on client node I get following problem: > > [root@node-a01 ~]# mount -t lustre storage01@tcp0:/datafs /mnt/datafs > mount.lustre: mount 10.143.245.1@tcp:/datafs at /mnt/datafs failed: > Input/output error > Is the MGS running? > > the IP address is OK and it is pingable from the client node. > > [root@storage01 ~]# ping storage01 > PING storage01.storage.cluster (10.143.245.1) 56(84) bytes of data. > 64 bytes from storage01.storage.cluster (10.143.245.1): icmp_seq=0 > ttl=64 time=0.022 ms > 64 bytes from storage01.storage.cluster (10.143.245.1): icmp_seq=1 > ttl=64 time=0.014 ms > 64 bytes from storage01.storage.cluster (10.143.245.1): icmp_seq=2 > ttl=64 time=0.013 ms > > Client node has LNET entry in modprobe.cfg for the right eth > interface. > > options lnet networks=tcp0(eth0) > > Why it isn''t working? > > cheers > > Wojciech Turek > > On 11 Jul 2007, at 21:15, Brian J. Murrell wrote: > >> On Wed, 2007-07-11 at 21:07 +0100, Wojciech Turek wrote: >>> Hi, >>> >>> I have a following command line: >>> >>> mkfs.lustre --fsname datafs --ost --mgsnode=storage01@tcp0 /dev/sdb >>> >>> Does --mgsnode above is correctly specified if MGS resides on the >>> same node that OST? >>> Host-name of the node is storage01 and modprobe.conf contains >>> following line: options lnet networks=tcp0(eth0) >> >> That looks correct. >> >> b. >> >> >> _______________________________________________ >> Lustre-discuss mailing list >> Lustre-discuss@clusterfs.com >> https://mail.clusterfs.com/mailman/listinfo/lustre-discuss > > _______________________________________________ > Lustre-discuss mailing list > Lustre-discuss@clusterfs.com > https://mail.clusterfs.com/mailman/listinfo/lustre-discuss