1325090 - Reduce number of HTTP_NET_VS_CACHE_ probes

Status: RESOLVED FIXED

(Core :: Networking: Cache, defect)

Description

[Michal Novotny [:michal]]

Several probes give more or less the same data, e.g. all HTTP_NET_VS_CACHE_ONSTART_ SMALL,MED,LARGE,HIGHPRI,NORMALPRI. We should revise all probes and keep only those providing useful data.

Comment 1

[Junior [inactive]]

Attachment: reduce_probes

I guess we will use the DiskStorageSize in the future, so I didn't remove it.
The threshold of the queue size depends on the priority, so I still leave the mCacheOpenWithPriority

Comment 2

[Michal Novotny [:michal]]

Comment on attachment 8826543 [details] [diff] [review]
reduce_probes

Review of attachment 8826543 [details] [diff] [review]:
-----------------------------------------------------------------

::: toolkit/components/telemetry/Histograms.json
@@ -1978,5 @@
> -      "kind": "linear",
> -      "high": 1000,
> -      "n_buckets": 100,
> -      "description": "Network vs cache time load (OnStopRequest) difference (ms) for cache files with a large size (>256K) and high priority. Offset by 500 ms."
> -  },

I didn't mean to drop them all. AFAICS in the telemetry there is not a big difference between each normal/high priority pair, so we can drop this flag completely. There is also not much difference between ONSTART time for different sizes, but this is not true for ONSTOP. So we can drop all those ONSTART because they are already covered by HTTP_NET_VS_CACHE_ONSTART_NOTREVALIDATED, but we should keep ONSTOP probes for different sizes. And because there is not a big difference between small and medium size we should either merge small and medium into a single probe, or shift the size limits.

Comment 3

[Junior [inactive]]

I wasn't aware it still need to record the onStopDiff for different sizes of content. My bad.

Comment 4

[Junior [inactive]]

Attachment: reduce_probes_v2

Comment 5

[Michal Novotny [:michal]]

Comment on attachment 8827782 [details] [diff] [review]
reduce_probes_v2

Review of attachment 8827782 [details] [diff] [review]:
-----------------------------------------------------------------

I'm not sure if we need data review when we're removing and updating probes.

Comment 6

[Benjamin Smedberg]

Comment on attachment 8827782 [details] [diff] [review]
reduce_probes_v2

removing no. Updating it's good to check. This is fine.

Comment 7

[Junior [inactive]]

(In reply to Benjamin Smedberg [:bsmedberg] from comment #6)
> Comment on attachment 8827782 [details] [diff] [review]
> reduce_probes_v2
> 
> removing no. Updating it's good to check. This is fine.

Actually we merged some, but it doesn't matter a lot.

Comment 8

[Junior [inactive]]

https://treeherder.mozilla.org/#/jobs?repo=try&revision=c322f9fe2237c999cb945b23b56978342a9758ed

Comment 9

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/586d65b40b24
Drop and merge HTTP_NET_VS_CACHE_* probes, r=michal

Comment 10

[Phil Ringnalda (:philor)]

https://hg.mozilla.org/mozilla-central/rev/586d65b40b24