Version
0.3.0
Environment
No response
Expected behavior
Description
We are building a vLEI system using KERIA. During a 48-hour long-run load test involving vLEI issuance and revocation, we observed that both CPU and RAM usage on the KERIA server continuously increase (leak) until they hit 100%.
Even when upgrading the AWS instance specifications, the increase only slows down but the pure upward trend remains unchanged. This behavior might be related to the escrow timeout or incomplete message deletion issues mentioned in #232.
Environment / Specs
- Cloud Provider: AWS (The issue occurs regardless of instance size, though larger specs slow down the accumulation rate)
Expected Behavior
Memory and CPU resources should be properly released after vLEI issuance/revocation operations are completed, preventing indefinite growth over long periods. Incomplete multisig escrows or processed messages should be deleted or timed out properly.
Actual behavior
- Resource Leak: Both CPU and RAM usage increase continuously during the test.
- CPU 100% Impact: Once the CPU usage reaches 100%, API calls and processing requests frequently fail with errors.
- RAM 100% Impact: Once the memory usage hits 100%, the KERIA process gets terminated (likely due to OOM / Out of Memory killer).
- After Restart: When the KERIA process automatically restarts, the CPU and RAM usage are reset back to normal levels.
Steps to reproduce
- Prerequisites: Passcodes for QVI and QVI vLEIs are already issued beforehand.
- Setup: Create 3 new LAR (Legal Accountability Representative) personas, including their passcodes.
- Execution (vLEI Operations):
- Issue the following vLEIs: LE, OOR Auth, OOR, ECR Auth, and ECR.
- Revoke all the issued vLEIs right after.
- Test Schedule: Run the above scenario within a 2-hour window, wait for a 2-hour idle interval, and then repeat it sequentially.
- Duration: Conduct this loop as a long-run test for 48 hours.
Version
0.3.0
Environment
No response
Expected behavior
Description
We are building a vLEI system using KERIA. During a 48-hour long-run load test involving vLEI issuance and revocation, we observed that both CPU and RAM usage on the KERIA server continuously increase (leak) until they hit 100%.
Even when upgrading the AWS instance specifications, the increase only slows down but the pure upward trend remains unchanged. This behavior might be related to the escrow timeout or incomplete message deletion issues mentioned in #232.
Environment / Specs
Expected Behavior
Memory and CPU resources should be properly released after vLEI issuance/revocation operations are completed, preventing indefinite growth over long periods. Incomplete multisig escrows or processed messages should be deleted or timed out properly.
Actual behavior
Steps to reproduce