Problems with Mageia 3 and 3.10.40 kernel ...

[ghmitch]

I am having serious problems after switching from kernel server 3.10.28 to kernel server 3.10.40. The pattern of the hard drive activity leds has changed radically with the upgrade on boot up and shut down, and when running, On 3.10.28 led activity is very smooth and deliberate, with 3.10.40 there is a lot of random flashing and flickering instead of sustained access. Multi-media performance has become a disaster with a lot of stuttering with virtually no CPU load and plenty of throughput to and from the gateway. I rebooted on the older kernel and everything runs uneventfully. Has anyone else out there noticed a problem like this?

[doktor5000]

You could run iotop as root to see what process(es) produce all that IO under the newer kernel.
Apart from that, what filesystems do you use?

[ghmitch]

Well, I was afraid you were going to ask that question. The answer as to file systems is a little bit of JFS plus Terabytes of btrfs, lots of it RAID 1. Btrfs generates an enormous number of processes which seemed to not cause any problems with multimedia streaming on 3.10.28, but all of a sudden with 3.10.40 something is tying up the system and it would not surprise me if it were indeed all those btrfs threads going on. What puzzles me is what could have changed between the two releases to cause such a problem. It is definitely not a load issue. It is more like a scheduler issue. I am using the server kernel because of 4GB RAM on 32bit Mageia. I looked over the change log when I did the update and did not spot anything particularly interesting although I certainly might have missed something. I will take a look at iotop when I get a chance to boot up on 3.10.40 again, but I am running gkrellm and didn't observe any particularly interesting io going on in terms of io (which I monitor closely) once the system was booted. Just an unending number of btrfs processes churning away. Altogether about 480 processes I believe. Thanks for answering my post, I was just hoping there was some pattern here that could be identified if others were having the same problem, but apparently not. - George