[LU-6619] lustre/obdclass does not get cleared while stopping/cleanup Lustre if setup had additional MDT present

Fri, 28 Feb 2020 00:04:00 +0000

Hi,

Pre-requisite to reproduce this bug :

1. A single Scientific Linux (release 6.6) VM, with min 1 GB memory and
50GB disk space.
2. A lustre setup 2.7.51 up and running on the above VM, in my case all
lustre components are configured on the same VM .
3. I have added 1 extra MDT of 15 GB to the lustre setup, the MDT was
created on loop device.
===================================

Steps to reproduce the issue :
===================================
1. run dd command to generate some IO on the lustre filesystem .
( dd if=/dev/zero of=/mnt/lustre/test bs=512M count=10).
2. Once IOs are completed , stop Lustre filesystem , i had executed
lustrecleanup.sh script (../lustre-release/lustre/tests/llmountcleanup.sh)
to unmount/stop the lustre.
3. After the unmount completes, lustre prints error message on the
terminal :module lustre/obdclass stil loaded
=======================================
Command prompt trace :
=============================
[root@localhost ~]# sh /var/lib/jenkins/jobs/Lustre-New-Test/workspace/default/lustre-release/lustre/tests/ llmountcleanup.sh
Stopping clients: localhost /mnt/lustre (opts:-f)
Stopping client localhost /mnt/lustre opts:-f
Stopping clients: localhost /mnt/lustre2 (opts:-f)
Stopping /mnt/mds1 (opts:-f) on localhost
Stopping /mnt/ost1 (opts:-f) on localhost
Stopping /mnt/ost2 (opts:-f) on localhost
2 UP mgc MGC192.168.102.13@tcp 9918b9be-ec01-ce40-5dc1-d4ebb297e839 5
3 UP mds MDS MDS_uuid 3
23 UP osd-ldiskfs lustre-MDT0001-osd lustre-MDT0001-osd_UUID 9
24 UP lod lustre-MDT0001-mdtlov lustre-MDT0001-mdtlov_UUID 4
25 UP mdt lustre-MDT0001 lustre-MDT0001_UUID 5
26 UP mdd lustre-MDD0001 lustre-MDD0001_UUID 4
27 UP osp lustre-MDT0000-osp-MDT0001 lustre-MDT0001-mdtlov_UUID 5
28 UP osp lustre-OST0000-osc-MDT0001 lustre-MDT0001-mdtlov_UUID 5
29 UP osp lustre-OST0001-osc-MDT0001 lustre-MDT0001-mdtlov_UUID 5
30 UP lwp lustre-MDT0000-lwp-MDT0001 lustre-MDT0000-lwp-MDT0001_UUID 5
Modules still loaded: ************
lustre/osp/osp.o lustre/lod/lod.o lustre/mdt/mdt.o lustre/mdd/mdd.o ldiskfs/ldiskfs.o lustre/quota/lquota.o lustre/lfsck/lfsck.o lustre/mgc/mgc.o lustre/fid/fid.o lustre/fld/fld.o lustre/ptlrpc/ptlrpc.o lustre/obdc lass/obdclass.o lnet/klnds/socklnd/ksocklnd.o lnet/lnet/lnet.o libcfs/libcfs/libcfs.o
==================================================
4. Now after this message, I had manually unmounted the additional MDT , which results into call traces.
5. Var/log/messages shows--> proc_dir_entry 'lustre/lov' already registered, proc_dir_entry 'lustre/osc' already registered followed by call trace .
5. After this , if lustre mount/start is attempted , then the process Hangs
sometime ,Again results into call traces at the backend.
6. Reboot is the solution to cleanup everything and start a fresh.
====================================================
Attaching /var/log/messages and dmesgs from the Lustre setup.
====================================================
Thanks,
Paramita Varma

Whamcloud Community JIRA

[LU-6619] lustre/obdclass does not get cleared while stopping/cleanup Lustre if setup had additional MDT present