#3057 HIGH Trial-3: Journal memory usage continues to increase over time

[Zarro Boogs per Child]

#3057: Journal memory usage continues to increase over time
-------------------------------+--------------------------------------------
  Reporter:  kimquirk          |       Owner:  tomeu  
      Type:  defect            |      Status:  new    
  Priority:  high              |   Milestone:  Trial-3
 Component:  journal-activity  |     Version:         
Resolution:                    |    Keywords:         
  Verified:  0                 |  
-------------------------------+--------------------------------------------
Comment (by tomeu):

 Replying to [ticket:3057 kimquirk]:
 > If you open a few activities and just let the XO stay up and running for
 a few days, the Journal takes more and more memory as displayed by the
 size of the donut ring.

 As far as I know, the Journal will do *nothing* when the laptop is idle.
 Unless some activity is saving every X time, but I don't know of any that
 does that right now. Even then, the memory should not go up.

 I think we should make very clear that the donut should not be used for
 debugging mem usage nor for entering bug reports (other than about the
 donut itself). The variable size of donut sector used by an activity is
 indicative of what portion of the *available memory* is using.

 In other words, if some process that is not an activity increases its
 memory usage, the sectors of every activity will increase.

 Words from danw:

 {{{
 <tomeu> danw: how much can we trust the donut in 558 regarding activity
 memory size?
  memory usage, I mean
 <danw> in sugar-jhbuild, not at all, because of swap
 <tomeu> danw: and on the xo?
 <danw> on the xo, it should give a good idea of the relative sizes of
 activities, but you can't use it for memory debugging
  and you can't even assume that if an activity changes size, that that
 means it's using a different amount of memory
 <danw> because, eg, if you join or leave the mesh, and dhcpd starts/stops,
 that will change the amount of free memory in the system, so the total ram
 represented by the donut changes, so the wedge sizes may change too
 <tomeu> danw: could be a non-activity process taking memory, true?
 <danw> right
 }}}

 The dev console tries quite hard to show an approx. real memory usage. We
 should use that value instead.

 Anyway, I'll leave a laptop switched on overnight and will monitor the
 general mem usage.

-- 
Ticket URL: <https://dev.laptop.org/ticket/3057#comment:1>
One Laptop Per Child <http://laptop.org/>