Idle CPU Usage

[PuffShr1k3]

Is it possible the idle CPU usage could be reduced please? Albeit the usage is low, but if sitting idle, could the CPU usage be brought down a bit - Other addons use 0% if not in use.

I have lengthened the intervals to attempt to mitigate polling being a culprit:

Nonetheless, thank you for this extension - Having fire and forget backups is a huge stress relief.

[bdraco]

cpu_time.zip

Here are py-spys for where the time is spent

[bdraco]

Its all the enumeration in the backup models. backups.models.backups

called from

    def nextBackup(self, now: datetime, include_pending=True):
        latest = max(filter(lambda s: not s.ignore() and (not s.isPending() or include_pending), self.backups.values()),
                     default=None, key=lambda s: s.date())
        if latest:
            latest = latest.date()
        return self._nextBackup(now, latest)

I think most of them could be changed to properties and use @cached_property than the overhead would likely disappear

[sabeechen]

There are some subtle problems caching those properties would cause.

I can make some little fixes that should cut the time used there in half, though its kind of mysterious that its spending so much time calculating .ignore(). Do you have a lot of backups?

I'll also add a new configuration option that the performance sensitive like yourself can use. Right now the addon checks once/second if it should make a backup so new backups always happen on time. I'll make this configurable so you can sacrifice a little scheduling accuracy to do less computation.

[bdraco]

At least on my systems:

system 1: 87 ignored backups. 4 non-ignored backups. system 2: 92 ignored backups. 4 non-ignored backups.

[PuffShr1k3]

I noticed I had 40 ignored backups, deleted older backups, so now I only have 9 ignored, which reduced the CPU usage from 0.4% to 0.3%. I did notice that Google Drive backup identified the changes in the amount of backups very quickly, which ties in with how the adon checks once a second. I'm happy to sacrifice scheduling accuracy for less computational overhead. Thank you for looking into this, let me know if I can help test, provide logs, etc.