Hi Henk,
I am debugging a case where vdbench journal verify is reporting corruption. Your input on this will be very useful.
Here are the steps.
1. Vdbench write + inline verify starts where anchor directory is nfs share and journal is on local ramdisk.
This was successful. Here are the logs.
https://www.dropbox.com/s/luhcabh2jkz4r0m/vdbench_write_phase_logs.gz?dl=0
2. Journal from ramdisk is copied to nfs share. No errors seen during copy.
3. At later time, vdbench journal verify reports corruption. vdbench was expecting version 'x' of write but found version 'x+1'.
07:58:35.668 localhost-0: ===> Data Validation Key miscompare.
07:58:35.983 localhost-0: The sector below was written Sunday, March 19, 2017 06:49:26.713 UTC
07:58:36.287 localhost-0: 0x000 00000000 0067b800 ........ ........ 00000000 0067b800 00054b0f cfd0cafc
07:58:36.713 localhost-0: 0x010 0e..0000 66736431 20202020 00000000 0fc40000 31647366 20202020 00000000
07:58:37.137 localhost-0: There are no mismatches in bytes 32-511
Logs for journal verify phase here.
https://www.dropbox.com/s/2iim4zi3s7zlgu7/vdbench_journal_verify_logs.gz?dl=0
This can happen in case where vdbench write was returned as failure but write was actually completed by storage target.
I looked at vdbench logs but couldn't find log which would indicate write failure.
Can you please let me know if file vdbench_data/vdb.1_1.dir/vdb.2_1.dir/vdb_f0003.file and offset 0x00000000 saw any write failures.
Thank you,
Karthick