YCSB workload A kills cockroach #20448
Description
BUG REPORT
I can consistently replicate Cockroach getting into a deadlock situation by running YCSB workload A
Workload A: Update heavy workload
This workload has a mix of 50/50 reads and writes. An application example is a session store recording recent actions.
This is how I run it:
Install
wget https://github.com/brianfrankcooper/YCSB/releases/download/0.12.0/ycsb-0.12.0.tar.gz
tar -zxvf ycsb-0.12.0.tar.gz
cd ycsb-0.12.0
wget https://jdbc.postgresql.org/download/postgresql-42.1.4.jar
mv postgresql-42.1.4.jar lib
ant
Create schema
DROP DATABASE IF EXISTS ycsb;
CREATE DATABASE ycsb;
CREATE TABLE ycsb.usertable (
YCSB_KEY VARCHAR (255) PRIMARY KEY,
FIELD0 TEXT,
FIELD1 TEXT,
FIELD2 TEXT,
FIELD3 TEXT,
FIELD4 TEXT,
FIELD5 TEXT,
FIELD6 TEXT,
FIELD7 TEXT,
FIELD8 TEXT,
FIELD9 TEXT
);
Load data
./bin/ycsb load jdbc -P workloads/workloada -p db.driver=org.postgresql.Driver -p db.url=jdbc:postgresql://cockroach-01:26257/ycsb -p db.user=root -p db.batchsize=1000 -p recordcount=50000
Run
for threads in 1 2 4 8 16 32 64 96 128 160 192; do
./bin/ycsb run jdbc -P workloads/workloada -p db.driver=org.postgresql.Driver -p db.url=jdbc:postgresql://cockroach-01:26257/ycsb -p db.user=root -p operationcount=0 -p maxexecutiontime=60 -s -threads $threads
done
On m5.xlarge with 500 GB gp2 storage (1500 IOPS with bursts up to 3000) cockroach gets into a deadlock at around after threads=96. Please note that running a single test with threads=96 is not sufficient, because the load accumulates with each run.
If deadlocks under a heavy load of updates is not considered a bug, please close the issue.