Opened 13 years ago

Closed 13 years ago

#1851 closed defect (wontfix)

100 CPU load after 4 days of operations operation #5225

Reported by: Madmaxx Owned by:
Priority: major Component: General
Severity: medium Keywords: 100 CPU LOAD
Cc: Sensitive: no

Description

After 1-2 days of operations actual #5197 on i686 and arm/xscale goes to 100 cpu load and looses the local SRG via-reader for unknown reason (=no log entry)

there is no log entry and reason why after several hours the normal decoding rises to 5000 msec timeout and holding the server on 100 cpu load. via decoding works only a few hours, and then goes from 1 second to another to 5000 msec/not found and hungs on this task.

i use ONLY newcamd, cccam is completly disabled, and it is ONLY the via-reader on 3,57 MHz. The reader itself is ok, tested with older versions from oscam 4500+ this problems never comes up.

I have this problem the last 200 trunk versions with the via reader.

Developers: a LONGER investigation with a local smartmouse with an srg via card is needed and a few hours up to 1-2 days of running, not just a quicktest!

My reader is 100% OK, i can reproduce this problem on different readers and on different platforms. It is possible. that a wrong ecm triggers the timeout problem. I use NO filter lists for this card.

But in my case, i see in logfile just from one moment to another, that sucessfully decoded ecm with 98 msec switches to 5000 msec/not found.
Loadbalancer is NOT in use/disabled.

Change History (6)

comment:1 by littlejoe, 13 years ago

Please try #5206 or later.

comment:2 by pooyair, 13 years ago

Resolution: wontfix
Status: newclosed

sorry , this info would not help dev to fix issue , you might provide valgrind log as explained here
http://www.streamboard.tv/wbb2/thread.php?postid=381036#post381036

comment:3 by Madmaxx, 13 years ago

Component: ReaderGeneral
Keywords: VIA removed
Resolution: wontfix
Severity: highmedium
Status: closedreopened

I let the #5225 Version running with unchanged config last days, NEWCAMD only with 9 local cards, and it was stable for 3 days. Now i got again a 100% cpu problem. Same situation: No information in logfile. But at the time of killing 1 thread was hanging and hard to kill.

It looks like that somewhere in the newcamd-code is not error-tollerant, maybe at the time of fast reconnect or get invalid ecm requests or something else.

There is absolutely no Information in the Logfile that show the trigger of the 100% load problem. The logfile just show normal ecm decodings and "top" shows the 100% load.

Without having a loggable trigger i have no information at which time the Highload is starting. Watching 100 hours by hand a "top" is useless to figure out the real reason for the high load. I need a better method.

Before 3 days oscam starts with 7.7 MB ram and ended now with 8.1 MB. So the 0.4 MB more ram after 3 days looks like the statistics data in ram and stored EMM cache from the last days. Nothing "unusual" here.

So this is a spurious bug, maybe buffer overflow, maybe memory leak that shows up after many hours. Maybe to fast reconnect with newcam account ?

comment:4 by Madmaxx, 13 years ago

Summary: 100 CPU load after 1-2 days of operations operation #5197100 CPU load after 4 days of operations operation #5225

comment:5 by pooyair, 13 years ago

@Madmaxx
this issue exists , but why don't u provide valgrind log to find where cause this!
also u could limit the min configs (ie disable lb other optional feature)... to find which part of oscam cause it .

Last edited 13 years ago by pooyair (previous) (diff)

comment:6 by john_28, 13 years ago

Resolution: wontfix
Status: reopenedclosed

provide log in ticket systems , or create a topic in forum and discuss your problem there ...

Note: See TracTickets for help on using tickets.