1248058 - Make sure that use of LOAD_BYPASS*_CACHE flags bypassing our DNS cache doesn't cause any performance issues

Status: NEW

(Core :: Networking: DNS, defect, P3)

Description

[Honza Bambas (:mayhemer)]

Bug 938803 introduced this.  Seems like LOAD_BYPASS_CACHE may be used frequently while it should not lead to DNS refresh every time.  One example is an XHR POST that should completely bypass caching (bug 1248049).  We don't necessarily need to use LOAD_BYPASS_CACHE for that case, but that is suboptimal: in case HTTP cache index is not up to date, we will on every XHR POST loop to disk to look for a file (if not found, no cache entry will be created), this is unnecessary IO.

We may use just the LOAD_BYPASS_LOCAL_CACHE to do DNS refreshes.

Comment 1

[Honza Bambas (:mayhemer)]

Err... we will ALWAYS loop to disk, regardless index status.  So, I think to fix bug 1248049 we need to pass LOAD_BYPASS_CACHE as well.  Which means a DNS refresh all the time.

Comment 2

[Patrick McManus [:mcmanus]]

I think the key thing here is to refresh the DNS cache when normal refresh is pressed. does LOCAL_CACHE do that?

Comment 3

[Honza Bambas (:mayhemer)]

(In reply to Patrick McManus [:mcmanus] from comment #2)
> I think the key thing here is to refresh the DNS cache when normal refresh
> is pressed. does LOCAL_CACHE do that?

It does.  But the correct question should be: does a normal refresh set BYPASS_LOCAL_CACHE?  No, it doesn't.

My consideration here is that we do refresh for literally every channel that is set BYPASS_CACHE or BYPASS_LOCAL_CACHE.  I think that should be made only for top level documents.  Otherwise it's pretty much waste of resources.

Comment 4

[Honza Bambas (:mayhemer)]

We should measure how often we do DNS refreshes that are not coming from F5.

Comment 5

[Patrick McManus [:mcmanus]]

honza, can you make a specific proposal?

Comment 6

[Honza Bambas (:mayhemer)]

(In reply to Patrick McManus [:mcmanus] from comment #5)
> honza, can you make a specific proposal?

let's talk about this live.  generally, I think some kind of a telemetry may give answers whether this is a problem.

Comment 7

[Firefox Bug Husbandry Bot]

Bulk change to priority: https://bugzilla.mozilla.org/show_bug.cgi?id=1399258

Comment 8

[Firefox Bug Husbandry Bot]

Bulk change to priority: https://bugzilla.mozilla.org/show_bug.cgi?id=1399258