FAHClient using entire core

absolutefunk

absolutefunk

New member
Howdy y'all!

When I woke up this morning I noticed that FAHClient was using an entire core, which was cutting my folding down to 75%. I rebooted which seemed to fix it, but I noticed tons of this in my log:

Server connection id=* on 0.0.0.0:36330 from 127.0.0.1

One entry with a new id every 10 seconds. I googled this and found a couple of other ppl having the same issue, one on Windows and one on MacOS, so this seems to not be a linux specifc problem. Anyone here seen this before? I can't seem to find a fix, and right now things are running fine.

-Brian
 
Just a quick update, those server connection messages in the log appear to be coming from FAHControl, which I had running during the night. It looks like at some point FAHControl went rogue and started establishing a new connection to the client every 10 seconds until it established 1012 connections (that's when the log stopped logging new connections). I think at that point FAHControl gave up, even though for each new connection there is no corresponding end connection, which tells me that FAHControl indeed had over a 1000 connections to the client before I rebooted. That would explain FAHClient using an entire core. I'm going to fold with FAHControl not running, and see what happens.

On a side note, FAHControl is the only monitoring program I have, so in theory I shouldn't get anymore phantom connection messages unless there is a strange bug in the client itself.

-Brian
 
Well it's not related to FAHControl. It seems that FAHClient goes rogue for some reason. Nothing in the log. I'm assuming FAHControl had the connection attempts earlier cause it couldn't physically connect to the client. Unfortunately if I can't figure this out then I won't be able to fold at all. Anyone have any insight into this?

-Brian
 
absolutefunk said:
Well it's not related to FAHControl. It seems that FAHClient goes rogue for some reason. Nothing in the log. I'm assuming FAHControl had the connection attempts earlier cause it couldn't physically connect to the client. Unfortunately if I can't figure this out then I won't be able to fold at all. Anyone have any insight into this?

-Brian
Click to expand...

I had a bad stick of ram a few months ago that under normal load (web and gaming) didn't cause a problem but for folding I got issues like you are reporting. I've also had my main folding rig do this but it was because I hadn't cleaned out the dust and it's was over heating.

I'm also wondering what version of the client you are running.
 
Bleeder said:
I had a bad stick of ram a few months ago that under normal load (web and gaming) didn't cause a problem but for folding I got issues like you are reporting. I've also had my main folding rig do this but it was because I hadn't cleaned out the dust and it's was over heating.

I'm also wondering what version of the client you are running.
Click to expand...

I'm running 7.2.9. I was wondering about hardware myself, but was thinking if it was hardware related then either Prime95 or the Intel Burn Test would have caught it. I think I figured it out. It dawned on me that I had two instances of the client running, which was strange as I had deleted the GPU slot (can't GPU fold in Linux). However I never had disabled the client for the GPU slot, and what I think happened is a bug related to a running client process which no longer has a slot. I reran the client using the default config (including GPU folding), and havn't had a problem since. Only annoyance is my logs get filled up with the 'empty assignment' query results for GPU WUs, but it backs off nicely after a while so not a big deal.

Btw, speaking of stress testing, I remember back in the day stress testing using Gromacs cores was typically heavier than using Prime, but I have noticed since folding that my core temps are about 5-10 degrees lower than what I was getting with Prime. Pretty nifty :)

-Brian
 
You must log in or register to reply here.
Back
Top