Closed ryderhsu0129 closed 2 years ago
@ryderhsu0129 : Could you share the parameters you were invoking the test with?
@yhr I just run the script zenfs_base_performance.sh, the parameters carried by the db_bench should be set by the script?
@ryderhsu0129 : right, if you are running the zenfs_base_performance.sh scripts the parameters are calculated by get_good_db_bench_params_for_zenfs.sh . Maybe, If you are testing on a device with huge zones(or one that reports wierd zone capacity values), the target file size base parameter calculation might cause the OOM. Hard to know for sure without knowing what you are testing on. As a reference, with zone capacity of ~1G, i get a memory usage of ~2G.
@yhr I have increased the memory capacity to 64GB, the capacity should be enough to finish the test. It has been tested for more than 24 hours, is this normal in general?If so how long does this test generally take? thanks.
@ryderhsu0129 : When I run it it normally takes about 8h or so, but it depends on the disk properties (write throughput, number of active zones)
@yhr OK, thanks for your reply.
Hi,
a basic db_bench performance testing we are running is causing our machines to run out of memory and get killed by the kernel.
[Test Log]
long_performance/0000_fillrandom.sh: line 7: 109458 Killed $TOOLS_DIR/db_bench $DB_BENCH_PARAMS >> $TEST_OUT
[dmesg]
Out of memory: Killed process 109458 (db_bench) total-vm:16862280kB, anon-rss:14872260kB, file-rss:0kB, shmem-rss:0kB, UID:0 pgtables:32036kB oom_score_adj:0
is there any advice to fix the issue?