[lustre-discuss] Lustre-2.15.X with ldiskfs cannot mount MGS/MDT

Wed Oct 12 02:09:11 PDT 2022

Hi,

I hit the same issue this weekend. 

>3. We compiled and installed e2fsprogs-1.46.2.wc5.
The /lib64/libext2fs.so.2 file is not installed by 'make install’ command but this library is used by mkfs.lustre

As workaround you can install e2fsprogs from a rpm.
Here is a link to el8 version for instance

https://downloads.whamcloud.com/public/e2fsprogs/latest/el8/RPMS/x86_64/

Best regards,
Artem Blagodarenko

> On 12 Oct 2022, at 03:38, Tung-Han Hsieh <thhsieh at twcp1.phys.ntu.edu.tw> wrote:
> 
> Dear All,
> 
> We manually compiled and installed Lustre-2.15.0 and 2.15.1,
> with backend file system ldiskfs. We found that it cannot mount
> the MGT/MDT partition at all (we formatted MGT and MDT in the
> same partition).
> 
> The symptom is following. After installation, we partitioned and
> formatted the Lustre file system using the following commands:
> 
> # mkfs.lustre --fsname ctest --mdt --mgs --index=0 /dev/sdb1
> # mkfs.lustre --fsname ctest --ost --mgsnode=<host>@tcp0 --index=0 /dev/sdb2
> 
> Then trying to mount /dev/sdb1, we got the error message:
> 
> =========================================
> # mount -t lustre /dev/sdb1
> mount.lustre: /dev/sdb1 has not been formatted with mkfs.lustre or the backend filesystem type is not supported by this tool
> =========================================
> 
> We further tried tunefs.lustre, and found that sdb2 seemed normal, but sdb1
> failed:
> 
> =========================================
> # tunefs.lustre /dev/sdb1
> checking for existing Lustre data: not found
> 
> tunefs.lustre FATAL: Device /dev/sdb1 has not been formatted with mkfs.lustre
> tunefs.lustre: exiting with 19 (No such device)
> 
> # /opt/lustre/sbin/tunefs.lustre /dev/sdb2
> checking for existing Lustre data: found
> 
>   Read previous values:
> Target:     ctest-OST0000
> Index:      0
> Lustre FS:  ctest
> Mount type: ldiskfs
> Flags:      0x2
>              (OST )
> Persistent mount opts: ,errors=remount-ro
> Parameters: mgsnode=192.168.50.3 at tcp
> 
>   Permanent disk data:
> Target:     ctest-OST0000
> Index:      0
> Lustre FS:  ctest
> Mount type: ldiskfs
> Flags:      0x2
>              (OST )
> Persistent mount opts: ,errors=remount-ro
> Parameters: mgsnode=192.168.50.3 at tcp
> 
> Writing CONFIGS/mountdata
> =========================================
> 
> Even if both partitions were pre-formatted and successfully mounted, stored
> data by Lustre-2.12.6 with ldiskfs, the MGT/MDT partition still cannot be
> mounted by Lustre-2.15.X.
> 
> On the other hand, there is no problem to mount MGT/MDT partition with
> ldiskfs in Lustre-2.15.X. If the partition is newly formatted, there is
> only CONFIG/mountdata with 12288 bytes. If it is pre-formatted and used
> by previous Lustre version, it contains.
> 
> =========================================
> BATCHID             lost+found/  oi.16.22  oi.16.4   oi.16.57
> CATALOGS            lov_objid    oi.16.23  oi.16.40  oi.16.58
> CONFIGS/            lov_objseq   oi.16.24  oi.16.41  oi.16.59
> LFSCK/              nodemap      oi.16.25  oi.16.42  oi.16.6
> NIDTBL_VERSIONS/    oi.16.0      oi.16.26  oi.16.43  oi.16.60
> O/                  oi.16.1      oi.16.27  oi.16.44  oi.16.61
> OI_scrub            oi.16.10     oi.16.28  oi.16.45  oi.16.62
> PENDING/            oi.16.11     oi.16.29  oi.16.46  oi.16.63
> REMOTE_PARENT_DIR/  oi.16.12     oi.16.3   oi.16.47  oi.16.7
> ROOT/               oi.16.13     oi.16.30  oi.16.48  oi.16.8
> changelog_catalog   oi.16.14     oi.16.31  oi.16.49  oi.16.9
> changelog_users     oi.16.15     oi.16.32  oi.16.5   quota_master/
> fld                 oi.16.16     oi.16.33  oi.16.50  quota_slave/
> hsm_actions         oi.16.17     oi.16.34  oi.16.51  reply_data
> index_backup/       oi.16.18     oi.16.35  oi.16.52  seq_ctl
> last_rcvd           oi.16.19     oi.16.36  oi.16.53  seq_srv
> lfsck_bookmark      oi.16.2      oi.16.37  oi.16.54  update_log
> lfsck_layout        oi.16.20     oi.16.38  oi.16.55  update_log_dir/
> lfsck_namespace     oi.16.21     oi.16.39  oi.16.56
> =========================================
> 
> We manually compiled Lustre-2.15.X with the following steps:
> 
> 1. Install vanilla Linux kernel kernel 5.4.135, because we saw:
> 
>   lustre-2.15.X/ldiskfs/kernel_patches/series/ldiskfs-5.4.136-ml.series
> 
>   The Linux kernel source was not patched by Lustre code.
> 
> 2. In lustre-2.15.X source tree, we run:
> 
>   ./configure --prefix=/opt/lustre --with-linux=/usr/src/linux-5.4.135 \
>               --with-o2ib=no --enable-mpitests=no
>   make
>   make install
>   depmod -a 5.4.136
>   export PATH="/opt/lustre/sbin:/opt/lustre/bin:$PATH"
>   export LD_LIBRARY_PATH="/opt/lustre/lib:$LD_LIBRARY_PATH"
>   ( cd /usr/sbin; ln -s /opt/lustre/sbin/ko2iblnd-probe . )
>   ( cd /usr/sbin; ln -s /opt/lustre/sbin/l_getidentity . )
>   ( cd /usr/sbin; ln -s /opt/lustre/sbin/l_tunedisk . )
>   ( cd /usr/sbin; ln -s /opt/lustre/sbin/lctl . )
> 
>   The contents of /etc/modprobe.d/lustre.conf is:
> 
>   options lnet networks="tcp0(eth0)"
> 
>   During the compilation and installation, there is no errors.
> 
> 3. We compiled and installed e2fsprogs-1.46.2.wc5.
> 
> 
> Could anyone see how to solve this problem ? Thank you very much.
> 
> Best Regards,
> 
> T.H.Hsieh
> _______________________________________________
> lustre-discuss mailing list
> lustre-discuss at lists.lustre.org
> http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org