Closed Bug 1502147 Opened 6 years ago Closed 6 years ago

Make pushes collapsible to free memory on completed pushes, when there still are running jobs below

Categories

(Tree Management :: Treeherder: Frontend, enhancement)

enhancement
Not set
normal

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: bogdan_tara, Assigned: camd)

References

(Depends on 1 open bug)

Details

Attachments

(2 files)

I think this would make the sheriffs lives easier:

Add a toggle to collapse a push, so that just the push date, e-mail, mercurial link and commit message are visible.

Ideally, the collapsed push would be discarded to free some memory on the sheriffs computers.

:camd is this worth a try?
Flags: needinfo?(cdawson)
Summary: TH Improvement: Make collapsible pushes to free memory on completed pushes, when there still are running jobs below → TH Improvement suggestion: Make collapsible pushes to free memory on completed pushes, when there still are running jobs below
Assignee: nobody → cdawson
Status: NEW → ASSIGNED
Flags: needinfo?(cdawson)
I think the issue this is wanting to resolve might be better fixed by bug 1107667 instead?
Attached a graphic sketch-up of my suggestion.
The idea is to free as much memory as possible and some screen space in cases when there still are running jobs below pushes with all tests done.
This happens quite often on autoland and inbound.
:emorley there are cases when a fixed page size would not help at all.
We already can select the top and bottom pushes - this can make TH faster when we're looking at 3-4 pushes.
This still feels like a premature optimization - we should first start by improving actual+perceived performance and memory usage, then after resort to these approaches if that still isn't enough.

In addition, hiding pushes retrospectively doesn't make the initial load faster - the pushes still have to load and then later manually be hidden one at a time. This seems like very manual-busywork for small performance gain.

I do agree that the performance on mozilla-central (where pushes have thousands and thousands of jobs) is particularly bad at the moment. Perhaps a better solution would be to set a limit on max jobs per push, and if exceeded, then hide passing builds by default or similar? Or default the page size to 5 not 10?
Summary: TH Improvement suggestion: Make collapsible pushes to free memory on completed pushes, when there still are running jobs below → Make pushes collapsible to free memory on completed pushes, when there still are running jobs below
One approach to improving performance might be only rendering the pushes that are in the current view-port. I've filed bug 1506062 for this :-)
If a push can be collapsed and kept collapsed on refresh/new tab, it could be helpful.
Bug 1506062 makes more sense. :)
While this certainly isn't the whole solution to the memory/performance issues, I still feel this is a worthwhile feature.  This impl is persisted in the URL for sharing.  It can be used to reduce noise in some of those circumstances.  

I'm really curious, though, how things like the package mentioned in bug 1506062 will work.  That could be really cool.

Tbh, though, I suspect there's something with how the pushes are stored and used with a Context.  Not certain yet, but I'm hunting.
Status: ASSIGNED → RESOLVED
Closed: 6 years ago
Resolution: --- → FIXED
This makes me so happy, thank you!
Depends on: 1511651
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: