Hi,
This rang a bell for me. I've had similar issues on my automon project using the ts7390. Think it was something to do with the sd-card driver, any time a write occured "the world would stop".
Have a look at
On Thu, Dec 6, 2012 at 9:10 PM, Rod <> wrote:
Les, thanks for your response. I forgot to mention that I also see this periodically at the shell prompt, where no characters are echoed for a few seconds while typing. The characters are not lost, simply delayed in being echoed.
This is becoming a significant concern.
--- rod.
--- In ts-7000%40yahoogroups.com, "aeroelastic" <> wrote:
>
> I see this as well, and am not using the CAN bus. I have killed most of the daemons running, apache, inetd, etc. still the pauses continue, and top does not reveal any culprit.
>
> It DOES sometimes seem like it is on the console or terminal (ssh) that the problem is the worst. Running application seem less prone to the dead spots.
>
> I have tried playing with the nice settings, but the problem persists on terminal and ssh
>
> Les
>
> --- In ts-7000%40yahoogroups.com, "Rod" <bomr@> wrote:
> >
> > I am seeing problems of applications running on a TS-7350 that go 'dead' (for lack of a better description) for periods of 2-5 seconds. This seems to happen irregularly, but generally every 1 to 10 minutes. I suspect that there is some daemon process running periodically that is hogging all userspace CPU, since running 'top' doesn't show anything, but does also appear to go dead for brief periods (although this is very difficult to assess for sure).
> >
> > I've turned off swapping, and have tried to eliminate anything else I thought was unnecessary for my application. There are no cron jobs running. Can anyone suggest where else to look?
> >
> > I'm using a default 2.6.21 kernel, and am also using the TS-CAN1 CANbus interface with the TS supplied driver. I suspect that the problem exists even without the CANbus, but the CANbus interface allows me to see the problem, as I am grabbing time-stamped messages from the bus, and I know those timestamps are accurate.
> >
> > Thanks.
> >
> > --- rod.
> >
>
__._,_.___ ![]()
__,_._,___ |