[macOS] 26GB of memory being used in the main process under gfx. GPU-process is not present.
Categories
(Core :: Graphics: Canvas2D, defect, P1)
Tracking
()
People
(Reporter: mojofawad, Unassigned)
References
(Blocks 2 open bugs, )
Details
Attachments
(3 files)
User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:127.0) Gecko/20100101 Firefox/127.0
Steps to reproduce:
I was just using dbdiagram.io to create a small db model
Actual results:
my OS (mac) told me I waas out of application memory and firefox was 26.10GB of that.
I don't know how "application memory" works, but I've only got 8GB of RAM haha
Expected results:
not this? I wish I could be more helpful.
Comment 1•1 year ago
|
||
The Bugbug bot thinks this bug should belong to the 'Core::Performance' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.
Comment 2•1 year ago
•
|
||
Few observations :
- All the memory is under the main process
- There is no gpu-process
- All the memory inder main process in in gfx.
@mojofawad :
- Can you type about:support in your browser, and paste its contents on this bug?
- I played around on the website but didnt see any memory increase. If possible, can you also share the db model you were working on? I can try to export it on the website and see if that leads to any memory increase.
- If you can consistently reproduce this bug, does it still reproduce if you set "gfx.canvas.accelerated = False" adn restart the browser?
Updated•1 year ago
|
(In reply to Mayank Bansal from comment #2)
Few observations :
- All the memory is under the main process
- There is no gpu-process
- All the memory inder main process in in gfx.
@mojofawad :
- Can you type about:support in your browser, and paste its contents on this bug?
- I played around on the website but didnt see any memory increase. If possible, can you also share the db model you were working on? I can try to export it on the website and see if that leads to any memory increase.
- If you can consistently reproduce this bug, does it still reproduce if you set "gfx.canvas.accelerated = False" adn restart the browser?
-
I’ll get you number 1 if I can.
-
It was 4 tables max with 2 columns per table. While I’m unable to reproduce the bug, I can explain what I was doing at the time:
I was trying to figure out how to add references (I didn’t read the docs at first.) I kept trying to write it similar to Mermaidjs and I kept getting an error. So I would try writing it different ways.
After posting this bug and closing out the browser, the memory usage went down to ~1.5GB and I was able to use dbdiagrams again.
- I can’t consistently reproduce it, unfortunately.
@Mayank Bansal
I'm not sure what you mean about pasting contents on this bug. I can copy my about:support
contents now, but not with the bug.
Update: I noticed creating new tables would cause the memory to go up.
I was able to get it up to 8GB using the following dbml:
table tablename1 {
column1 int
}
table tablename2 {
column1 int
column2 int
}
table tablename3 {
column1 int
column2 int
column3 int
}
table tablename4 {
column1 int
column2 int
column3 int
column4 int
}
table tablename5 {
column1 int
column2 int
column3 int
column4 int
column5 int
}
ref {
"tablename1"."column1" > "tablename2"."column1"
}
ref {
tablename2.column2 > tablename3.column2
}
ref {
tablename3.column3 > tablename4.column3
}
Steps to reproduce:
- Don't copy/paste this. Start in the default view and type each table in one by one.
- Delete the dbml that was just written, and without refreshing, enter this in the dbml, one table at a time
table a {
a int
b datetime
}
table b {
b datetime
c somevalue
}
table c {
c bool
d char
}
ref {
a.b > b.b
}
ref {
b.c > c.d
}
Behavior I noticed:
- upon deleting the original dbml, the memory usage did not go down
- memory usage increases with each new table written. this behavior isn't replicated when copying 20+ tables all at once
Comment 7•1 year ago
•
|
||
Thanks, I can reproduce now. I need to type-in only till table3, and then with each new variable added in table3, memory starts increasing rapidly.
Gpu-canvas: I get ~1.5GB of memory in gpu-process when i type out the the first three tables. https://share.firefox.dev/3XTtFjd
D2d-canvs: 9.5GB used at the system level (but only 300mb in any of the Firefox processes).
Can repro on a build from Jan2021, so not a new regression.
Maybe the site is increasing the height and width of the canvas each time you type a new table+column name which leads to OOM?
Comment 8•1 year ago
|
||
Updated•1 year ago
|
Comment 9•1 year ago
|
||
With gpu-canvas:
Profile with memory tracking : https://share.firefox.dev/4bdrr0R
Profile with graphics preset logging: https://share.firefox.dev/4biKSW9
Comment 10•1 year ago
|
||
Lee, do you know about how this might be happening?
Comment 11•1 year ago
|
||
I can reproduce this as well.
Comment 12•1 year ago
|
||
I think it's ok to keep this as S2 until we learn what's going on.
Updated•1 year ago
|
Comment 13•1 year ago
|
||
(In reply to Erich Gubler [:ErichDonGubler] from comment #10)
Lee, do you know about how this might be happening?
Not sure offhand. Maybe related to bug 1904048.
Comment 14•1 year ago
•
|
||
If it is related to bug 1904048, then maybe OP (mojofawad) can test on Nightly, since it's resolved?
Updated•1 year ago
|
Updated•1 year ago
|
Comment 15•1 year ago
|
||
Myank, can you see if you can reproduce this based on comment 7?
Comment 16•1 year ago
|
||
(In reply to Bob Hood [:bhood] from comment #15)
Myank, can you see if you can reproduce this based on comment 7?
I can no longer reproduce this.
Updated•1 year ago
|
Updated•1 year ago
|
Description
•