# on the server #> lctl get_param *.*.recovery_status obdfilter.myth-OST0004.recovery_status= status: COMPLETE recovery_start: 1594074852 recovery_duration: 278 completed_clients: 1/3 replayed_requests: 0 last_transno: 120259116149 # abort recovery lctl --device myth-OST0004 abort_recovery # Get OSC device list from the client lctl dl # Recover a device lctl --device <device number or name> recover # on the MGS: get the llog lctl --device MGS llog_print $fsname-client lctl --device MGS llog_print $fsname-OSTxxxx lctl --device MGS llog_print $fsnmae-MDTxxxx # evicting client lctl set_param mdt.*.evict_client=<uuid> lctl set_param mdt.<target>.evict_client=uuid:<client_uuid>" or "...=nid:<client_nid> #or lctl set_param mdt.*evict_client=nid:<nid> # setting to decide whether "mdt.*evict_client=N" is only on the local MDT or on all OSTs as well lctl set_param mdt.*.evict_tgt_nid=[0|1]" lctl set_param mdt.$fsname-MDT0000.evict_tgt_nids=<> #get the UUID lctl get_param llite.*.uuid # You can mount the device which is used for a lustre server target and inspect the configuration there mount -t ldiskfs /dev/vdb /mnt/tmp # get stripe information for directory lfs getstrip -d * # trim file system. run on all servers fstrim -a #drop all the locks on the client lctl set_param ldlm.namespaces.*.lru_size=clear # enable fake IO. Good for performance testing lctl set_param fail_loc=0x238
Overview
Content Tools