Since upgrading to FileMaker Server 16 our solution is suffering from server-side scripts triggered by the Perform Script On Server being occasionally interrupted without any apparent reason and without any meaningful record in any log.
The weird thing is that the same script runs fine maybe hundred times, but once, randomly, the script session gets interrupted. What could be the reason? What are all possible cases for a server-side script sesssion to get interrupted without any error or warning in the Event.log?
I previously learned that server-side scripts launched via PSOS with waiting for result get interrupted when the calling script gets interrupted or the client gets disconnected. But this is happening to scripts launched via PSOS without waiting for result and the calling script continues running just fine on the client.
I temporarily enabled Top Call Stats with 1 second interval to collect as much data as possible and were able to spot this issue while having the Tap Call Stats on. I have made logs and screenshots documenting the issue available here: Dropbox - PSOSinterruptions.zip
The "all_in_one.txt" file is a text file with data from all logs consolidated into one.
Screenshot "interrupted_script.png" shows the script that got interrupted, screenshot "suspect_log_records.png" shows the exact log records at the time the script got interrupted.
We use FM Bench to benchmark all our scripts. FM Bench uses ExecuteSQL via plug-in to insert records in the log. As shown in the screenshots and logs, the interrupted script was supposed to create 2 log entries at its very beginning, but it only created one. Immediately after it the session got interrupted and started closing databases. Our "onLastWindowClose" trigger got triggered, but it reported Error 1 (user cancelled error according to docs), and did not perform. Then all databases got closed.
Any ideas why this might be happening? Is it a bug in FileMaker Server?
We're currently running FileMaker Server 184.108.40.2064 on macOS Sierra 10.12.6 (16G1114). I have not tried the 16v4 update yet but the releases notes seem like there's no fix or change related to this issue.