Skip to contentSkip to conversation

perf/001

PATCH Anand Jain 2 participants 2 messages Oct 13, 2021 – Oct 14, 2021 Btrfs file system

lore
Share & tools
lore

Anand Jain

author
#0 lore raw

perf/001 fails to update the SQL with the key trim_bw_bytes.

Traceback (most recent call last):
File "/xfstests-dev/src/perf/fio-insert-and-compare.py", line 31, in
<module>
result_data.insert_result(data)
File "/xfstests-dev/src/perf/ResultData.py", line 45, in insert_result

    self._insert_obj('fio_jobs', job)

File "/xfstests-dev/src/perf/ResultData.py", line 37, in _insert_obj
cur.execute(cmd, tuple(values))
sqlite3.OperationalError: table fio_jobs has no column named trim_bw_bytes

I tried to add the missing keys (which are probably present only in the current version of the fio) as below, but the error is still the same. Any inputs, how to fix this?

src/perf/fio-results.sql +3 -0
index 62e1464834b0..5d115764c4e9 100644
@@ -8,6 +8,9 @@ CREATE TABLE IF NOT EXISTS `fio_runs` (
 CREATE TABLE IF NOT EXISTS `fio_jobs` (
   `id` INTEGER PRIMARY KEY AUTOINCREMENT,
   `run_id` int NOT NULL,
+  `trim_bw_bytes` int,
+  `read_bw_bytes` int,
+  `write_bw_bytes` int,
   `read_bw_dev` float,
   `trim_lat_ns_mean` float,
   `read_runtime` int,

Thanks, Anand

#1 lore raw
Quoted 8 lines · Anand Jain

perf/001 fails to update the SQL with the key trim_bw_bytes.

Traceback (most recent call last):
File "/xfstests-dev/src/perf/fio-insert-and-compare.py", line 31, in
<module>
result_data.insert_result(data)
File "/xfstests-dev/src/perf/ResultData.py", line 45, in insert_result

Quoted 9 lines
    self._insert_obj('fio_jobs', job)

File "/xfstests-dev/src/perf/ResultData.py", line 37, in _insert_obj
cur.execute(cmd, tuple(values))
sqlite3.OperationalError: table fio_jobs has no column named trim_bw_bytes

I tried to add the missing keys (which are probably present only in the current version of the fio) as below, but the error is still the same. Any inputs, how to fix this?

I think this was a failed experiment to add performance testing to xfstests, I think we should just rip it out. Thanks,

Josef