Memory Usage Increases heavily on packet sending

[Prinkesh]

After using guppy to profile heap usage i found that most of the memory which is not freed are instances of engineio.packet.Packet

from guppy import hpy
hp = hpy()
after= hp.heap()

 (Pdb) after
 Partition of a set of 238362 objects. Total size = 399495864 bytes.
  Index  Count   %     Size   % Cumulative  % Kind (class / dict of class)
      0   4077   2 369537464  93 369537464  93 unicode
      1  90861  38  9798800   2 379336264  95 str
      2  51605  22  5016536   1 384352800  96 tuple
      3    784   0  2264704   1 386617504  97 dict of module
      4  13381   6  1712768   0 388330272  97 types.CodeType
      5    875   0  1681928   0 390012200  98 dict (no owner)
      6   1763   1  1599856   0 391612056  98 type
      7   1763   1  1583432   0 393195488  98 dict of type
      8  13164   6  1579680   0 394775168  99 function
      9  35057  15   841368   0 395616536  99 int
 

 (Pdb) after.byrcs[0]
 Partition of a set of 2373 objects. Total size = 368690936 bytes.
  Index  Count   %     Size   % Cumulative  % Referrers by Kind (class / dict of class)
      0   2373 100 368690936 100 368690936 100 dict of engineio.packet.Packet
 
 (Pdb) i = iter(after.byrcs[0].nodes)
 
 (Pdb) i.next()
 
       Shows data of the packets which were sent by the server
 

What could be the problem ?

Could this happen if the clients are slow in receiving the packets ?

[miguelgrinberg]

What did your server and client do while these measurements were taken? Any code you can share to help me reproduce here?

[DeoLeung]

notice the same behaviour

using python==3.7.2 tornado==5.1.1, redis manager

it increases after a js client disconnect (meanwhile a python write only client keep generating data from server)

I suspect that would be the AsyncRedisManager, seems it lacks a proper closing method

[miguelgrinberg]

This is extremely old, and memory leak bugs have been addressed since then, so I assume it is fixed.