Having 100 publishes purposefully started concurrently causes about 15% of the jobs to fail. Each one of these publishes instantiated another 5-6 Javelin processes.
When this issue was investigated, the following was determined:
- Parallel requests corrupt the Javelin logs if they start in the same second.
- This is because the file name ends with the second the javelin flow was created.
- One of the job hence gets orphaned and the portal status remains RUNNING as well.
- The problem seems to be due to two jobs writing into one javelin file, created in the same second.
- The file name may be the problem and including the millisecond may solve the issue.
Based on the above, Javelin was changed to not allow files to be named the same.
Javelin-126.96.36.199.zip has been supplied as the fix for this issue. Please contact Support for a fix if you are experiencing this issue.