With the suggestion of @JayKay729, I decided to create a dummy backup for testing how Kopia choose which backups to save, but in this experiment, I learned much more than I initially planned. In this post, I would like to share these discoveries, along with some examples to demonstrate them. This is a supplement to his post, so I’ll jump straight to those he hasn’t covered yet.
Numbers in Snapshot Retention
Except for Latest Snapshots, Kopia considers the number you write here as duration, not amount. For example, in this image, 48 is not 48 hourly snapshots, but rather, retain 1 snapshot per hour within 48 hours from your most recent snapshot, while 7 is retain 1 snapshot per day within 7 days from your most recent snapshot, with the following results:
Despite listing 48 hourly snapshots, only 3 snapshots were tagged as daily as only these 3 fit the 48-hour duration of 5/25/2024 11:00:00 AM to 5/27/2024 11:00:00 AM. The same applies to 7 daily snapshots, with only 4 snapshots fitting within the 7-day duration of 5/20/2024 to 5/27/2024.
Or in other words, only the most recent snapshots within the duration you specified will be considered for tagging. Anything that is older than the duration you specified will be ignored.
Intervals of Each Snapshot Retention
Have you ever wonder what Kopia considers as a week, is it Sun-Sat, Mon-Sun, or 7 days from the most recent snapshot? Or have been befuddled by Kopia producing more than 1 daily snapshot? Stay tuned, as I break them down in this section.
Hourly Snapshots
Interval = hh:00:00 - hh:59:59
The snapshot tagged with hourly is the latest snapshot within that hour. Once the snapshot exceeds even a second from that hour, it would get a separate hourly tag, like in this example:
where as soon as the clock struck 12:00:00 AM, it is considered another hour, and hourly-3 tag was assigned to 11:55:34 PM, matching @jkowalski’s response before, and most users intuition.
Daily Snapshots
Interval = UTC 12:00:00 AM - 11:59:59 PM?
Based from my tests, it does not follow my system’s time. Neither is it the server’s time since these are local backups, directly on my PC’s storage. Instead, I presume it follows a standard time like UTC, possibly to avoid conflicts if the source and destination are on different time zones (common in online backups). Here’s an example:
Here, snapshots going as far as 4:55 AM of the next day are still considered to be within that day, which matches my timezone. And this is why you can get more than 1 daily snapshot: because they’re on different days in UTC.
Weekly Snapshots
Interval = Monday to Sunday, but only checking day, not time
Checking day? Yes, it looks only at the day, not the time, or in other words, follows system date. Confusing? Check this example:
Notice here that the weekly tag is applied on the latest snapshot of the week of May 20 to 26, where as soon as the date changed to May 27, it’s already considered another week. This is in stark contrast to daily that follows UTC instead, resulting in a mismatch (unless you live in UTC-0).
This means that the gap between daily and weekly tags will differ based on where you live, ranging from exact match (UTC-0) to 14 hours difference (UTC+14). Other tags like monthly and annual possibly follow the same scheme, looking at the month or year, respectively.
Interaction Between the Numbers and Intervals of Snapshot Retention
What happens if you don’t rely on latest snapshots and let the other retention policies choose which snapshots to keep? Fortunately, the behavior is what most users expect: as long as you hit the end of an interval, a tag will be assigned, as this example demonstrates:
3rd Snapshot
4th Snapshot
As 3:00:05 PM exceeded the interval for hourly snapshot (2:59:59 PM), the previous snapshot (2:59:07 PM) gets the hourly-2 tag, even if the gap between the oldest and most recent snapshot is less than an hour (which also means that this old issue has been resolved in the more recent Kopia versions). Continuing further on 3 more snapshots:
7th Snapshot
shows that the hourly tag properly protects the snapshot from the previous hour from getting deleted.
Disabling Snapshot Retention
What if you don’t need a certain retention policy? Put “0” on it, like this:
As shown in this pic, all hourly tags disappeared.
DON’T just leave them blank, as they would revert to the previous values you last saved:
Let me know if you have further questions, or have spotted some errors. Feel free to share your opinions too, on whether you find this helpful, or find some parts needing clarification.
I may take a first crack at it & then ask for your review when you return.