Auto-Schedule-Restore Broken... unless you reboot on a regular basis

Black Hole

May contain traces of nut
The backups listed in the WebIf backup tab (which are also made automatically, as you say) and stored in /mod/var/backup/ are separate from the one true auto-schedule-restore backup, which is /mod/boot/schedule.ab.

The auto-schedule-restore backup is created soon after the system starts up, by the script for which I posted an update in the the package thread linked above.

Today I auto-retuned one of my machines (a HD-FOX as it happens, in order to grab Python on That's TV), then rebooted to restore the schedule. Sure enough it did a second reboot, and "restored" some kind of schedule... but at least weeks out of date and apparently not comparable with any auto or manual backup in my list.

So I did the same on a HDR-FOX to track the process, and got a similar result. The problem has been side-stepped by manually restoring from the most recent auto-backup (in both cases).

What you have to realise is that I (for one) do not reboot unless there is a reason (fault or maintenance). I had no idea the auto-restore was a separate backup only made at boot time, but recalled something had been mentioned about the subject not long ago and went searching for it.

Only making an auto-restore backup at boot seems short-sighted. What is the rationale? Why can't auto-restore simply use the most recent non-empty auto (or manual) backup? Why isn't the auto-restore backup not exposed in the backups list where we can see it?
 
See df's quote in Post 1
Oh yes. I just checked the times on ours which confirms it's at boot.

I also notice that it only does one a day - I often set up recordings in the morning and on those days it does a backup then and the evening start doesn't then create an entry.
 
I guess that ASR was coded before the schedule backup. Once it works OK I have no problem with it operating as it does. I'd rather apply the backup manually. invalid backups can be made if the DBs have been corrupted while the settop program carries on with its in-memory copy, and the next restart (actually the one after) is when ASR typically gets invoked.
 
I'd rather apply the backup manually.
Anyone who'd rather apply backups manually can simply not install auto-schedule-restore.

The purpose of ASR is to minimise the impact of an unwanted (and possibly unattended) network-initiated auto-retune or random factory reset, and the impact won't be minimum if the backup ASR works from is significantly out of date – which it is likely to be for 24/7 units. There is a defect in that argument in that the auto-restore requires manual reboot...
 
So it looks like I need to promote implementing a daily reboot, not only to take advantage of ASR but also tunefix-update and for crash recovery.
 
Last edited:
I don't (immediately) see why ASR couldn't do a daily backup. tunefix-update will force a reboot as and when necessary if the machine is not busy, so you don't need to worry about that.
 
Back
Top