hyperledger
[BUG] CPU leak during load testing
OS and Environment
k8s, linux
GIT commit hash
2ffcd007e3f794f8ee32488a0446bc11c1242cd0
Minimum working example / Steps to reproduce
run the job in Jenkins https://jenkins.soramitsu.co.jp/job/iroha2/job/deploy/job/iroha2-perf/ with the following parameters
testScenario maxPerf
logLevel INFO
configuration standard
concurrentUsers 2
times 7200
deachLevelLasting 10
separatedByRampsLasting 10
startingFrom 0
leave other parameters unchanged
iroha2 config recommended
"COMMIT_TIME_LIMIT_MS": 10000,
"MAX_TRANSACTIONS_IN_BLOCK": 20,
"ACTOR_CHANNEL_CAPACITY": 100,
"GOSSIP_BATCH_SIZE": 500
"GOSSIP_PERIOD_MS": 1000
},
"BLOCK_SYNC": {
"GOSSIP_PERIOD_MS": 1000,
"BLOCK_BATCH_SIZE": 4,
"ACTOR_CHANNEL_CAPACITY": 100
},
analytical info
Actual result
- when running the script
- the load should increase evenly
- transactions should be accepted by the computer
- no CPU leaks should occur
Expected result
When the load is applied, a leak in the CPU is observed, judging by the logs from the profiler, the problem is related to the regular calculation of the hash of blocks and transactions. Analytical information can be found at the link to the profiler
logs downloaded from the parca profile.pb.gz
Logs in JSON format
Log contents
look at the profiler logs
Who can help to reproduce?
@RamilMus
Notes
No response
the problem persists when assembling Linux-based Alpine, the problem is solved when assembling Image based on Debian