View Full Version : Linux client issues on MP systems

04-09-2007, 06:27 PM
I'm having a strange issue with the latest linux cient v2.9013-498-CFR-06082120, but only on MP systems. After the client runs for a few days it loses one of the threads. One thread simply goes idle and requires a stop and restart of dnetc. I've seen the issue on three different versions of Linux. On the hardware side there are the new Core 2 Duo Xeons, Core Duo lv Xeons and old school P4 DP Xeons. On the DP Quad core Xeon system only one thread goes idle leaving 7 threads processing. All the other systems are Dual thread systems so they drop one thread with one thread still running. Has anyone else seen this behavior or know of a fix other then restarting the client every few days?

04-10-2007, 04:19 AM
By "goes idle", do you mean that the process goes down to 0% CPU usage, or that it disappears entirely? Do you run your clients hidden (-hide), or can you see what's happening from the front-end? Does it still toggle between showing #a and #b on the duallies?

The only thing I can say is that my FreeBSD SMP box has been untouched for months at a time in the recent past with no problems.

04-10-2007, 12:28 PM
The process is still there but it drops to 0% cpu usage. From what I see dnetc starts up and does a quick CPU check then spawns extra threads as necessary to load down each cpu/core. It's the initial thread that is going idle.

04-11-2007, 04:25 AM
check this topic and see if this also works on Linux clients


also do you transmit directly or do you buffer??

frequent-threshold-checks=2 <---- with 2 you transmit directly , i know if you do not change this setting and you have no internet connection, OGR wil kill the tread and dont restart again like you pointed out above

let me know if it works :thumbs: