Well, this should have been titled "heavy IO hangs system".
I obviously mislead most readers.
Most responders thought either:
1) my problem was NFS paging.
2) the problem was memory
1) no NFS activity is occuring
2) Larger memory will reduce paging rates, but these are IO rates
masquerading as paging rates. Larger memory will just defer where
this occurs, I believe. My complaint is the system locks up, instead
of degrading more or less gracefully.
The most promising response was:
From: email@example.com (Steven M Cohn)
There is a mode in the Berkeley paging code known as "panic mode"
If real memory becomes sufficiently low, the system will literally
stop doing everything and start freeing up memory. Several minutes
sounds like a long time, but anything is possible.
This seems to be what is occurring; the disk light comes on about once
every 3 or 4 seconds, and then only briefly. This occurs after paging
rate as measured by vmstat, or rstatd goes over 1000.
thanks to all,
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:06:14 CDT