Main INDEX
Monthly INDEX
PREV
NEXT
User name R. Michaels
Log entry time 04:04:34 on February12,2000
Entry number 33623
keyword=cpu states when ER1/deadtime problems
The two times tonight we got "ER1 disconnect" which after a
couple minutes led to largish deadtime fluctuations we
noticed the CPU became ~0% idle and ~80-90% bound
in "iowait", though there were no processes in the process
list that dominated. Once when we stopped ESPACE (on adaqL1)
the deadtime sank from the 50% to normal 10%. So far, marginal
evidence that it helps to run ESPACE on the previous run,
not the present run, when running at or above ~1 kHz. Jumping
disk head ? Maybe I'm wrong. Keep an eye on it.