[LU-11531] group descriptors corrupted! ldiskfs_check_descriptors: Checksum for group 23024 failed (37296!=369)

• Type: Bug
• Resolution: Not a Bug
• Priority: Critical
• Severity: 3
• Assignee: WC Triage
• Reporter: Artem Blagodarenko (Inactive)

Description

During test bluepil03 failed to start:

2018-09-28 04:24:58 [  108.843382] LDISKFS-fs (md66): ldiskfs_check_descriptors: Checksum for group 23024 failed (37296!=369)
2018-09-28 04:24:58 [  108.843382] LDISKFS-fs (md66): group descriptors corrupted!

test: 6 clients, IOR && mdtest,

DSTRIPECOUNT=4 MDTINDEXRAND=true DELAY=120 MPIRUN=/usr/bin/mpiexec.hydra MPIRUN_OPTIONS="--wdir=/home/mpiuser -iface ib0 " \
ha_nonmpi_loads="" ha_mpi_loads="ior mdtest" IORP='"-f /storage/shared/bluepill/grev/fpp"' \
MDTESTP='"-T -F -r -C -u -L -V 1 -i 20 -I 50 -z 1 -b 10"' \
ha_mpi_instances=1 STRIPEPARAMS="-c -1" \
mpi_threads_per_client=16 ha_load_timeout=120 \
POWER_DOWN="pdsh -S -w bluepill00 pm -0 " POWER_UP="pdsh -S -w bluepill00 pm -1 " \
FAILBACK="pdsh -S -w bluepill00 cscli failback -n " \
/usr/lib64/lustre/tests/ha.sh  -s bluepill02,bluepill03,bluepill04,bluepill05,bluepill06,bluepill07,bluepill08,bluepill09 -d /mnt/fs1 \
-c bluepill-client03,bluepill-client04,bluepill-client05,bluepill-client06,bluepill-client09,bluepill-client10 \
-u 36000  -p 120 -v bluepill02,bluepill03