Play nice question

[themeister]

I'm using Donator 0.81.056 running on a quad-core server using SageTV with SageDCT capturing data from HDHR Prime tuners. I use SageTV Extenders to render to several TVs in the house. I have a GbE wired network throughout. I capture to a 4-disk RAID-5 array as well as other SATA drives.

I have set play_nice_sleep=10 and thread_count=1. I start 4 copies of comskip (working on 4 shows). I start comskip as soon as the recording starts.

I notice that comskip will use 100% of a processor even with play_nice. I don't want to eat all 4 processors, because this is a server that performs other things; however, I do want to have it operating on more shows simultaneously. If I could control its processor use, I would run 5 at a time, because I have 5 tuners.

This is because frequently (especially with sporting events), I will have 4 or more games recording while I am watching one. I want to stay behind "live" so that the commercials get removed. I switch back and forth between games.

With this version of comskip, I can get flawed recordings either because of processor or disk contention, either one of which is caused by comskip being too aggressive.

Is there a solution to slow comskip down? Thanks.

--Bill

[erik]

Make sure you also set the play nice start and end times.
Otherwise nothing will change.

Example to have play nice active from 8:00 till 23:00
play_nice_start=0800
play_nice_end=2300

[themeister]

That worked...

Thanks, Eric.

[themeister]

Now I'm getting "pulsing". The spikes are causing pulsing in the playback (pauses at a very regular interval) as well as corruption in recordings. This version is just too aggressive in the use of system resources to co-exist with a lot of capture and playback going on. I didn't have this problem with earlier versions.

What would be a recommended earlier version to go to that uses less system resource?

Thanks,
Bill

[erik]

CPU access is not the problem as Comskip runs at idle priority
Its most certain that filling the read buffers over the network causes the problem.
Not sure how to prevent that

[themeister]

I didn't say it was CPU. Is it possible that this version does more I/O than prior versions. I've gone years without a problem. I upgraded and now I have a problem. I'm not saying it's even comskip's problem.

All I ask is that you recommend a prior stable version that I could try to see if the problem goes away.

Thanks.

--Bill

[erik]

The latest versions use larger read buffers to reduce disk trashing in multi comskip configurations.
This however seems to create a network load problem in your situation.
The readme.txt should note when the read buffer is increased