Quote:
Originally Posted by Red Squirrel
Hmm so if one raid array is busy it ties up the whole system?
|
Maybe - depends on how the disks are configured. Is this just a "standard" PC with IO busses on the motherboard ?. Or are you using iSCSI, fibre channel, something else ... ?
Quote:
So any iowait means the system is waiting on IO?
|
No, it doesn't - it means the system is idle, and there is uncompleted I/O still oustanding. Subtle but important difference. Have a read of
this, it seems a reasonable explanation.
Quote:
Is there a way to see which files are being accessed in real time?
|
iotop works like top, but for I/O - however it show what processes are doing the most I/O, not which files are being hit. I/O is done by disk (sector) address - I know of no easy way to do the "reverse lookup" to work out the file(s) involved.
As suggested, collectl gives finer data, and also has history on a per-process basis, so you can do analysis for bad periods in the past.
Another tool of interest may by latencytop, but CentOS may not have the required kernel options enabled for that to work.