touch-packages team mailing list archive

[James Hunt <1198180@xxxxxxxxxxxxxxxxxx>]

@uj__:

> The init process is certainly leaking memory.
How have you proved that init is leaking memory? Can you say that PID 1 is not just still busy processing the flurry of events that would result from starting all these docker containers? Please provide further details.

> I've looked at the code and it seems like there may be some leaks.
Where?

> The upstart tests are are also leaking memory. Some of that memory is leaked in the tests themselves (allocated and never freed)
The tests may well leak memory as they are not designed to be long-running like upstart.

> and some others seem to be leaking in libnih itself.
Again, where? Please provide logs. Note that libnih goes to some lengths to avoid leaks by doing automatic garbage collection. If you can produce a test program demonstrating a leak, please attach to this bug so we can investigate further.

-- 
You received this bug notification because you are a member of Ubuntu
Touch seeded packages, which is subscribed to upstart in Ubuntu.
https://bugs.launchpad.net/bugs/1198180

Title:
  possible leak in upstart 1.5

Status in “upstart” package in Ubuntu:
  Confirmed

Bug description:
  Description:    Ubuntu 12.04.2 LTS
  Release:        12.04
  Package:    1.5-0ubuntu7.2
  Kernel:      Linux XXX 3.8.0-23-generic #34~precise1-Ubuntu SMP Wed May 29 21:12:31 UTC 2013 x86_64 x86_64 x86_64 GNU/Linux

  We're starting lxc containers on demand, so there's lots and lots of
  network interfaces going up and down. Udev emits signal, that goes to
  upstart-udev-bridge, and next to the upstart itself. After about 40K
  containers created/destroyed, upstart takes about 250MB or RAM, and it
  grows with every event received/send. Not only RAM is beeing used, but
  also init spends more time for every new event on the socket.

  What happens:
  init uses more RAM and more CPU during lifetime, when lots of network interfaces are beeing created/destroyed

  What is expected:
  init work should not change depending on number of events served

  technical detail:
  # pmap -x 1
  00007f446f865000       0  237036  237036 rw---    [ anon ]
  00007fffc694d000       0      24      24 rw---    [ stack ]
  00007fffc69fe000       0       4       0 r-x--    [ anon ]
  ffffffffff600000       0       0       0 r-x--    [ anon ]

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu/+source/upstart/+bug/1198180/+subscriptions