AnsweredAssumed Answered

FMS 13 Aborted server Side Script stays connected, may cause 8003 errors and duplicates instances

Question asked by VincentL on Jul 21, 2014
Latest reply on Dec 3, 2014 by VincentL

Summary

FMS 13 Aborted server Side Script stays connected, may cause 8003 errors and duplicates instances

Product

FileMaker Server

Version

13v2

Operating system version

Mavericks 10.9.3

Description of the issue

During the night I've a Scheduled server side scheduled called "Queue Monitor Night". It runs every 10 minutes. It's basically a script that's a launcher of other server side enabled scripts. If there's a script due, then it executes it. So the launched script, is launched under the "Queue Monitor Night".

Those sub scripts can be long, up to 1 hours. Of course, while scripts are run, "Queue Monitor Night" is running, and hence can't be relaunched by the admin scheduler. I've set the "Queue Monitor Night" schedule to be limited to 2h30 running time.

Tonight, at 3 AM, "Queue Monitor Night" was launched at 5:34 AM it was aborted by the server according to the 2h30 limit.

Then, 10 min after, the schedule "Queue Monitor Night" was launched again at 5h44, which is normal BUT the previous "Queue Monitor Night" client, the one that should have been aborted, was still here.

That Schedule took a lot of time and then was aborted 2h30 later at 8h24.

Then 10 minutes later, at 8h34 it was launched again which is normal, but the 2 other instances even though they were supposed to be aborted were still connected !

So, when I checked out this morning at 8h37, 3 instances of the same scheduled were connected !

I'm pretty sure at least the slowness of the second instances is due to the fact the previous one was still connected and hence locking some file.

At 8 O'clock, first automated backup of the day (none during the night). Fails because 8003 error on a file (used by the server sides scripts), so server sends me warning mail, that's how I was warned.

So something made first schedule take way to much time (I'm thinking about table lock but can't see it in log),  then script aborted but in fact is not, so schedules instances pilles up cause server thinks the scheduled is aborted while is not.

Steps to reproduce the problem

See above

Expected result

Aborted script by server should REALLY disconnect the schedule client

Actual result

The server aborted the schedule, thinks it's aborted while it's not.

Configuration information

By the way, I tried to disconnect the "Queue Monitor Night" schedule instances with the admin, it was impossible (while I was able to disconnect the other users).

I then tried with fmasadmin command line, impossible to disconnect.

I then tried to close all database, but those uses by the "Queue Monitor Night" instances remained opened no matter what, even with command line.

I had to stop the server with command line and restart

Workaround

None

Photo_Partage_d%E2%80%99%C3%A9cran%2C_21_juillet_2014_08.37.16_UTC%2B2.png

Outcomes