On Mon, 11 Oct 2004 12:36:29 -0700, Johnny Walton <[EMAIL PROTECTED]> wrote: > > > I'm using plain ole mv. The virtual machine is 512M > and the data is a single file that is 650M. The > source nad destination are on seperate volumes. What > I'm seeing in iostat is nothing being written or read > "0 rkB/S 0wkB/s" for quite some time big await's then > I see a flurry of activity and await goes down but > starts getting larger and larger because nothing is > being written. Then the cycle starts again.
I have some concerns about whether the numbers from iostat are accurate in timing. But the behaviour you mention looks like all memory filled with data read from disk, and waiting for more memory to be freed up. But that happens only when the dirty pages are flushed out. As long as there is no huge amount of dirty data, it will flush small amounts every 5 seconds. Maybe you could verify that with /proc/meminfo? Could you give it a try and lower some of the limits in /proc/sys/vm/bdflush (e.g. the percentage at which the flush goes into continuous mode). I do have some plumbing that we could use to digest TRSOURCE and/or CP TRACE output to see what I/O took how long and when they were issued, etc. But that is heavy work. > I'll let you borrow my redbooks anyday;-) I have my own signed softcopy ;-) -- Rob van der Heij rvdheij @ gmail.com ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
