On Tue, 2012-02-21 at 11:03 +0100, Vincent Lefevre wrote: > In order to try to find the cause of the problem, I've set the atop > interval to 1 second. It seems that when the problem occurs, there > doesn't seem to be a particular activity of the processes (no > additional RDDSK or WRDSK). Just when the disk becomes busy, atop > shows that the "slab" value suddenly increases, then the disk > becomes busy for several seconds with writes only. During this > time, the "slab" value decreases. [...] In your example, the 'slab' number (kernel heap) only grows by about 4 MB and that doesn't seem to be directly forcing writeback since the 'buff' number (dirty buffers) doesn't get any smaller. There is a net increase of 5 processes over this period (186 -> 191). Maybe the extra disk activity comes from loading programs that aren't in cache? Ben. -- Ben Hutchings If at first you don't succeed, you're doing about average.
Attachment:
signature.asc
Description: This is a digitally signed message part