Discussion:
[omniORB] Servant do nothing, but get about 100% CPU
Sergey Kholodilov
2009-09-14 15:22:19 UTC
Permalink
Hello,

No clients, no connections, servant just waiting and do nothing. But
sometimes, may be after some days or weeks it starts eating a lot of CPU.

FreeBSD 7.0-STABLE
omniORB 4.1.3 (it beter than 4.1.2. 4.1.2 have this problem too, and in
4.1.2 it happens more frequently)
gcc 4.2.1
Amd64 (I hope, it is not matter, but may be...)

ktrace looks like:

94319 WMCDispatch CALL _umtx_op(0xb125a0,0x8,0x1,0xb12580,0x7fffff9fdd90)
94319 WMCDispatch CSW stop kernel
94319 WMCDispatch CSW resume kernel
94319 WMCDispatch RET _umtx_op -1 errno 60 Operation timed out
94319 WMCDispatch CALL clock_gettime(0,0x7fffff9fddf0)
94319 WMCDispatch RET clock_gettime 0
94319 WMCDispatch CALL clock_gettime(0,0x7fffff9fdda0)
94319 WMCDispatch RET clock_gettime 0
94319 WMCDispatch CALL _umtx_op(0xb125a0,0x8,0x1,0xb12580,0x7fffff9fdd90)
94319 WMCDispatch CSW stop kernel
94319 WMCDispatch CSW resume kernel
94319 WMCDispatch RET _umtx_op -1 errno 60 Operation timed out
94319 WMCDispatch CALL clock_gettime(0,0x7fffff9fddf0)
94319 WMCDispatch RET clock_gettime 0
94319 WMCDispatch CALL clock_gettime(0,0x7fffff9fdda0)
94319 WMCDispatch RET clock_gettime 0
94319 WMCDispatch CALL _umtx_op(0xb125a0,0x8,0x1,0xb12580,0x7fffff9fdd90)
....

There are no FreeBSD port for omniORB 4.1.4, so I don't try new version yet.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.omniorb-support.com/pipermail/omniorb-list/attachments/20090914/9d550fa9/attachment.htm
Michael
2009-09-14 20:28:00 UTC
Permalink
Hello Sergey,

most probably this is related to a bug in FreeBSD 7.0 which was fixed in
7.1 and 7.2, so please upgrade (it is an exception related multi
threading race condition - something goes wrong when unwinding the
stack). You can trigger this easily (I have some example code around
here, but basically just throwing an exception in some parallel remote
upcalls is enough).

Ever since we upgraded these problems disappeared (forgot which kernel
file has been patched for this though), but this problem was exclusive
to 7.0 (neither 6.x nor 7.1ff are affected)

cheers
Michael
Post by Sergey Kholodilov
Hello,
No clients, no connections, servant just waiting and do nothing. But
sometimes, may be after some days or weeks it starts eating a lot of CPU.
FreeBSD 7.0-STABLE
omniORB 4.1.3 (it beter than 4.1.2. 4.1.2 have this problem too, and in
4.1.2 it happens more frequently)
gcc 4.2.1
Amd64 (I hope, it is not matter, but may be...)
94319 WMCDispatch CALL _umtx_op(0xb125a0,0x8,0x1,0xb12580,0x7fffff9fdd90)
94319 WMCDispatch CSW stop kernel
94319 WMCDispatch CSW resume kernel
94319 WMCDispatch RET _umtx_op -1 errno 60 Operation timed out
94319 WMCDispatch CALL clock_gettime(0,0x7fffff9fddf0)
94319 WMCDispatch RET clock_gettime 0
94319 WMCDispatch CALL clock_gettime(0,0x7fffff9fdda0)
94319 WMCDispatch RET clock_gettime 0
94319 WMCDispatch CALL _umtx_op(0xb125a0,0x8,0x1,0xb12580,0x7fffff9fdd90)
94319 WMCDispatch CSW stop kernel
94319 WMCDispatch CSW resume kernel
94319 WMCDispatch RET _umtx_op -1 errno 60 Operation timed out
94319 WMCDispatch CALL clock_gettime(0,0x7fffff9fddf0)
94319 WMCDispatch RET clock_gettime 0
94319 WMCDispatch CALL clock_gettime(0,0x7fffff9fdda0)
94319 WMCDispatch RET clock_gettime 0
94319 WMCDispatch CALL _umtx_op(0xb125a0,0x8,0x1,0xb12580,0x7fffff9fdd90)
....
There are no FreeBSD port for omniORB 4.1.4, so I don't try new version yet.
------------------------------------------------------------------------
_______________________________________________
omniORB-list mailing list
http://www.omniorb-support.com/mailman/listinfo/omniorb-list
Sergey Kholodilov
2009-09-15 15:31:05 UTC
Permalink
Michael,

thank you for answer, your version looks very probably. It's unbelivable,
but right tonight our system administrators upgraded servers to FreeBSD 7.2
:) I will watch servants work, may be problem will gone.
--
Sergey
Post by Michael
Hello Sergey,
most probably this is related to a bug in FreeBSD 7.0 which was fixed in
7.1 and 7.2, so please upgrade (it is an exception related multi
threading race condition - something goes wrong when unwinding the
stack). You can trigger this easily (I have some example code around
here, but basically just throwing an exception in some parallel remote
upcalls is enough).
Ever since we upgraded these problems disappeared (forgot which kernel
file has been patched for this though), but this problem was exclusive
to 7.0 (neither 6.x nor 7.1ff are affected)
cheers
Michael
Post by Sergey Kholodilov
Hello,
No clients, no connections, servant just waiting and do nothing. But
sometimes, may be after some days or weeks it starts eating a lot of CPU.
FreeBSD 7.0-STABLE
omniORB 4.1.3 (it beter than 4.1.2. 4.1.2 have this problem too, and in
4.1.2 it happens more frequently)
gcc 4.2.1
Amd64 (I hope, it is not matter, but may be...)
94319 WMCDispatch CALL
_umtx_op(0xb125a0,0x8,0x1,0xb12580,0x7fffff9fdd90)
Post by Sergey Kholodilov
94319 WMCDispatch CSW stop kernel
94319 WMCDispatch CSW resume kernel
94319 WMCDispatch RET _umtx_op -1 errno 60 Operation timed out
94319 WMCDispatch CALL clock_gettime(0,0x7fffff9fddf0)
94319 WMCDispatch RET clock_gettime 0
94319 WMCDispatch CALL clock_gettime(0,0x7fffff9fdda0)
94319 WMCDispatch RET clock_gettime 0
94319 WMCDispatch CALL
_umtx_op(0xb125a0,0x8,0x1,0xb12580,0x7fffff9fdd90)
Post by Sergey Kholodilov
94319 WMCDispatch CSW stop kernel
94319 WMCDispatch CSW resume kernel
94319 WMCDispatch RET _umtx_op -1 errno 60 Operation timed out
94319 WMCDispatch CALL clock_gettime(0,0x7fffff9fddf0)
94319 WMCDispatch RET clock_gettime 0
94319 WMCDispatch CALL clock_gettime(0,0x7fffff9fdda0)
94319 WMCDispatch RET clock_gettime 0
94319 WMCDispatch CALL
_umtx_op(0xb125a0,0x8,0x1,0xb12580,0x7fffff9fdd90)
Post by Sergey Kholodilov
....
There are no FreeBSD port for omniORB 4.1.4, so I don't try new version
yet.
Post by Sergey Kholodilov
------------------------------------------------------------------------
_______________________________________________
omniORB-list mailing list
http://www.omniorb-support.com/mailman/listinfo/omniorb-list
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.omniorb-support.com/pipermail/omniorb-list/attachments/20090915/89ac4a36/attachment.htm
Loading...