I seem to be having a problem with CentOS releases 4.2 and 4.3. On a fresh install (on two very different pieces of hardware, one Dell PowerEdge server and one Compaq Presario), the system seems to be extremely slow. It happens primarily during any hard disk I/O.
It's slow enough that my mouse skips slowly instead of moving smoothly, and I am unable to type. It seems to freeze for milliseconds at a time, which makes typing difficult because while it is frozen it seems to not accept keystrokes (that is, if I type the letter 'f' twelve times, only four of them may appear, depending on timing). It may be unrelated, but the clock seems to go more slowly during I/O as well (i.e. during times of I/O the clock runs behind more and more). If I were to extract a large tarball, I would have to wait for the while thing to finish before I could interact with the system at all
A simple recompile of the kernel from kernel.org sources (which, as you can imagine, takes quite a long time) fixes the problem entirely, though. So is there something that differs wildly regarding the I/O scheduler between the stock CentOS kernel and the stock kernel from kernel.org?
It would really be more convenient to be able to keep our kernel up- to-date automatically using the built-in packaging tools, so I hope that someone has seen the problem and has a work-around
Just one more piece of information: a simple recompile isn't what has fixed the problem every time, the recompile has involved an upgrade to 2.6.16 as well
On 21 Aug 2006, at 15:06, David King wrote:
I seem to be having a problem with CentOS releases 4.2 and 4.3. On a fresh install (on two very different pieces of hardware, one Dell PowerEdge server and one Compaq Presario), the system seems to be extremely slow. It happens primarily during any hard disk I/O.
It's slow enough that my mouse skips slowly instead of moving smoothly, and I am unable to type. It seems to freeze for milliseconds at a time, which makes typing difficult because while it is frozen it seems to not accept keystrokes (that is, if I type the letter 'f' twelve times, only four of them may appear, depending on timing). It may be unrelated, but the clock seems to go more slowly during I/O as well (i.e. during times of I/O the clock runs behind more and more). If I were to extract a large tarball, I would have to wait for the while thing to finish before I could interact with the system at all
A simple recompile of the kernel from kernel.org sources (which, as you can imagine, takes quite a long time) fixes the problem entirely, though. So is there something that differs wildly regarding the I/O scheduler between the stock CentOS kernel and the stock kernel from kernel.org?
It would really be more convenient to be able to keep our kernel up- to-date automatically using the built-in packaging tools, so I hope that someone has seen the problem and has a work-around _______________________________________________ CentOS mailing list CentOS@centos.org http://lists.centos.org/mailman/listinfo/centos
-- David King Computer Programmer Ketralnis Systems
David King wrote:
Just one more piece of information: a simple recompile isn't what has fixed the problem every time, the recompile has involved an upgrade to 2.6.16 as well
There is a bad patch to the vm that makes the box use swap even when there is free RAM space available. Maybe that is what you are seeing.
The fix will be in the 4.4 release kernel.
Just one more piece of information: a simple recompile isn't what has fixed the problem every time, the recompile has involved an upgrade to 2.6.16 as well
There is a bad patch to the vm that makes the box use swap even when there is free RAM space available. Maybe that is what you are seeing.
I don't think so, swap usage is only 440k
The fix will be in the 4.4 release kernel.
What kernel version will be in 4.4? I can look this up myself too if someone can kindly give me a pointer :)
On Tue, 2006-08-22 at 11:52 -0700, David King wrote:
Just one more piece of information: a simple recompile isn't what has fixed the problem every time, the recompile has involved an upgrade to 2.6.16 as well
There is a bad patch to the vm that makes the box use swap even when there is free RAM space available. Maybe that is what you are seeing.
I don't think so, swap usage is only 440k
The fix will be in the 4.4 release kernel.
What kernel version will be in 4.4? I can look this up myself too if someone can kindly give me a pointer :) _______________________________________________
2.6.9-42.EL (on the 4.4 CD when released)
or
2.6.9-42.0.2.EL (new out today ... will be an update)
-
David King -
Feizhou -
Johnny Hughes