Consider resolved queries that take less than X ms as "resolved by cache"

[valantur]

I don't use AdGuard's cache and rely on a local Unbound recursive DNS server to do all caching.
However, I love AdGuard's other features and its UI and make decisions based on the information presented there. Therefore, is in my best interest that the info I see on AdGuard's UI is as consistent and accurate as possible.

Right now, if I want to know if a certain query was resolved by my resolver's cache or not, I have to check if the response time was less than 1ms. That tells me my AdGuard instance asked Unbound and Unbound replied from cache.

It would be nice to have an option named "Consider responses that take less than X ms to be returned from cache". This will ensure that Adguard reports accurate information. Even better, you could create a new category of response called something like "Returned from local upstream servers cache" that would let users differentiate between what system's cache is being used, but at the same time consider these responses as cached responses instead of forwarded responses.

Sorry for the wall of text... I hope this makes sense to the Devs.

Thanks!

[fernvenue]

Hi @valantur, first, make sure you are talking about AdGuardHome, AdGuardHome and AdGuard are totally different things. Second, AdGuardHome already tells us that these requests are coming from cache or not in response details:

And actually, you can disable cache by setting cache_size to 0, hope this can help you :)

[cyb3rz3us]

Actually calculators do give "wrong" answers all the time such as whenever one uses Pi or the natural log. Both are non-repeating numbers and consequently, the calculator uses an approximation. The same is also true with any number of trig functions.

My point is this: presumptions, approximations, etc. are perfectly reasonable provided all of the interested parties are aware that the presumptions, approximations, etc. are being used and the limitations of same.

Thank you for making my point...

[fernvenue]

the calculator provides an approximation

Do you really think the exact opposite conclusion is an approximation? In the case we are discussing, AdGuardHome may give "resolved by cache" when the upstream server using recursive lookup, vice versa.

[cyb3rz3us]

Seriously man - this has gone on long enough. The endless contrarian view with zero substantive rationale serves no useful purpose. The case for the feature has been made - and made quite well - and it's now up to the devs...

[fernvenue]

First of all, I opened this topic in AdGuardHome's repo. Not AdGuardForWindows or any other.

I forgot this, some people really don't know the difference between them, like #4712 this one. So I'm just make sure you are using AdGuardHome.

[fernvenue]

The endless contrarian view with zero substantive rationale serves no useful purpose.

I've given enough probabilities that the truth could be quite the opposite, let me quote it again for you:

So when the upstream server response time is greater than the time you specified, how can you be sure that the response must not come from the cache? Couldn't it be caused by network fluctuations? Couldn't it be that it took more time to establish the connection this time? Couldn't it be that the server is overloaded and the response is slower this time? And you called this option "resolved by cache", how do you ensure that this doesn't mislead users when AdGuardHome having no definitive way to determine its conclusion?

And if you are talking about yourself, then yes, I really don't see you giving any substantial reason why it's reasonable to tell users the exact opposite conclusion.

[cyb3rz3us]

telling the user a guess is misleading the user.

Actually, as long as the feature is well defined and documented, it isn't misleading at all. If a user decides to not RTFM & misunderstands the feature, that's not being misled - that's willful ignorance. Can you imagine if we opted to not build features because some users didn't RTFM? We'd never have any features...

As an aside, it is puzzling (to me at least) that @fernvenue is in favor of filtering query responses based on response time but is against having a feature akin to the following:

---

External cache tracking

• Consider & log all query responses equal to or below X milliseconds as served from external cache.

[fernvenue]

Actually, as long as the feature is well defined and documented, it isn't misleading at all.

I agree with you, but "resolved by cache" is clearly not well-defined, "resolved by cache" is too much, it must be fact based, not a possibility.

[cyb3rz3us]

it must be fact based

We can agree to disagree here but regardless, that would be a question for the devs. Are they OK with having a feature that presents data based on a clearly understood presumption or only fact based. Both methods are reasonable depending on the situation.

[fernvenue]

Both methods are reasonable depending on the situation.

So tell me yes or no: "resolved by cache" is well-defined or not? If your answer is no, why "...Both methods are reasonable..." ?

[ainar-g]

I don't really feel like this is a feature that should be added to AdGuard Home. Low query durations do not necessarily mean that they were resolved from the upstream cache. And conversely, query durations above the threshold don't mean that the queries weren't resolved from cache. I feel like this kind of statistic is better off in a custom query log database, which will be possible once #3389 is resolved.

[fernvenue]

Totally agree.