Two nights ago, I ran the scheduler for an automated all night capture sequence of M33.
Everything went without a hitch, but in the morning, as it was getting light, I noticed that the scheduler was still running and happily exposing frame after frame of the morning sky.
I double-checked all the settings, the twilight box was ticked.
That never happened before, the scheduler in the past would always and reliably terminate and park the mount when astronomical twilight time was reached.

Here the relevant screenshots and link to the extensive log file for that night.

www.dropbox.com/sh/em3bfkd6vpo6j79/AACOw...OZ6sBNct4XKYFEa?dl=0

I can confirm this issue as well. The last two nights I have been imaging M31 and M33 and each morning the system was still running exposures well past twilight. My settings were also on as El Corazon shows. I may have logs if needed, however I will not be able to check that until later today. I am running the nightly build as updated two days back on a Tinkerboard.

There is a new feature for handling aborted jobs. As it can be seen from the screenshot, „reschedule immediately“ is selected. Unfortunately, the version you use contains a bug handling constraints leading to a restart although constraints are not met.

It is fixed meanwhile and available when compiling from sources.

As a workaround, set the error handling to „Don‘t re-schedule“.

- Wolfgang

Danke, Wolfgang,

but I don't think that's it.

I had suspected that this might be a part of the problem and played around with that setting as the exposures continued into dawn.

The scheduler actually ran over twilight while set on the default setting (I believe that's 'Don't reschedule'). I just forgot to set it back from 'Reschedule immediately' before taking the screenshot.

Also, the scheduler had correct switched to the next target in line. So everything was working well as it should, except for the twilight shutoff.

I'll give it another try, though, and make sure the setting is to not reschedule.

In which update has the bug been fixed? Nightly version (sorry, just reread your instruction: compiling from sources)?

Jo

Hi Jo,
yes, compiling from sources. Do you have logs? If not, maybe you could try to reproduce it with simulators.

Wolfgang

The log file is included in the Dropbox link in the first post. It is very large.

Hi Jo,
in the log file it looks like the second job NGC281 has no twilight limitation, since it is scheduled for 05:45 to 06:54, as could also be seen in your first screenshot. If it has been left unmodified since, you could upload your scheduler file here for a check.

- Wolfgang

Hi Wolfgang,

Yes, that may be part of the problem. The two jobs, M33 and NGC281, were set up in the evening and started together, with the twilight setting on in the scheduler. I never worried about the finish time of the job being later than twilight, since I figured the calculations would just state how long it would take to acquire all the images and then the scheduler would always reschedule after the twilight shutoff.

There is another problem. When you look at the scheduled and completed M33 sequence, it shows start time and end time as 22:21 and run as completed. The run completed alright, all 240 frames were acquired, but it took all night until ~5.45 am when NGC281 started after M33 was finished (look at the log file contained in the Dropbox link). So at least the end time in the scheduler is incorrectly stated.

Since I don't know how the scheduler is coded, I don't know whether that could impact on its failure to shut down at twilight.

Herzliche Grüsse,

Jo

No, that should not have an impact, since the scheduler checks the constraints after each captured frame. I'll screen the log file and will come back.

I took a closer look into your capture time for your first target

• M33 Ha (90x120sec): 22:24 - 01:37 - 13 min on top of capturing
• M33 OIII (90x60sec): 01:37 - 03:36 - 29 min on top
• M33 SII (60x120sec): 03:36 - 05:54 - 18 min on top

So instead of the calculated 24354 secs, your session took roughly 27000 secs. That's the reason why your session took until 05:54 and not 05:10.

Yes, that is all clear. The extra time is readily explained by focusing, dithering, downloading images from the camera buffer, meridian flip and realignment.
But it does not explain why the scheduler shows that M33 started at 22;21 and ended during the same minute, i.e. also at 22:21, and not at 5:45, when it actually ended.
That must still be a bug in the scheduler.

Wolfgang, I think this comment might describe the same problem with the twilight setting:

www.indilib.org/forum/ekos/5757-feedback...ion-3-3-6.html#44283