Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 8 additions & 0 deletions src/sqlancer/voltdb/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -421,6 +421,14 @@ was built with clang's source-based coverage, with the JaCoCo agent on every ser
coverage per Java package, per engine directory and file, and per area (planner, executors, DR, export, rejoin and
so on).

The runs' intervals follow `--duration` in the nightly job's proportions (its 2,700 s runs kill a node every 300 s and
crash the cluster every 600 s): a kill every `duration / 9` s (at least 60) and a crash every `duration / 4` s (at
least 120); the security run uses TLS (`--ssl`). Until milestone 44 the intervals were fixed, and milestone 43's 10-minute
runs made one kill and no crash at all. Measured again with crashes (milestone 44: the crash and security runs, 10
minutes each, master bf8602e3), command log and recovery went from 46.5% of their lines to 58.7%
(`CommandLogReplayState` 359 of 512 lines, from 88; `CommandLogReinitiatorImpl` 207 of 270, from 95; `ReplaySequencer`
66 of 97, from 7), and TLS setup (`SslSetup`) to 141 of 348 lines, from 18.

Build the engine with coverage in a clean checkout of VoltDB (macOS; Linux with clang works the same way):

```
Expand Down
13 changes: 9 additions & 4 deletions voltdb/run-coverage.sh
Original file line number Diff line number Diff line change
Expand Up @@ -169,17 +169,22 @@ if ! $report_only; then
done
fi
i=0
# the kill, export and crash runs' intervals follow --duration as the nightly job's do (2,700 s: a kill every 300 s,
# a crash every 600 s), so a short run still kills and crashes: with fixed intervals, milestone 43's 600 s crash run
# made no crash at all, and recovery from the command log went unmeasured
kill_every=$((duration / 9 > 60 ? duration / 9 : 60))
crash_every=$((duration / 4 > 120 ? duration / 4 : 120))
IFS=, read -ra run_list <<<"$runs"
for run in "${run_list[@]}"; do
s=$((seed + i))
i=$((i + 1))
case $run in
security) args=(--nodes 1 --kfactor 0 --sites 8 --security) ;;
kill) args=(--nodes 3 --kfactor 1 --sites 8 --kill-interval 300) ;;
export) args=(--nodes 3 --kfactor 2 --sites 2 --export --kill-interval 300) ;;
security) args=(--nodes 1 --kfactor 0 --sites 8 --security --ssl) ;;
kill) args=(--nodes 3 --kfactor 1 --sites 8 --kill-interval "$kill_every") ;;
export) args=(--nodes 3 --kfactor 2 --sites 2 --export --kill-interval "$kill_every") ;;
resize) args=(--nodes 3 --kfactor 1 --sites 8 --resize-interval 120) ;;
xdcr) args=(--nodes 3 --kfactor 1 --sites 4 --xdcr) ;;
crash) args=(--nodes 3 --kfactor 1 --sites 8 --crash-interval 600) ;;
crash) args=(--nodes 3 --kfactor 1 --sites 8 --crash-interval "$crash_every") ;;
repros | sqlcoverage) args=() ;;
*) die "unknown run $run (see --runs)" ;;
esac
Expand Down