sorry to chime in so late but I only just saw this thread.
As mark said, you should try out collectl, but even more important you might consider installing colmux as well, which is part of collectl-utils.
Whenever I have questions about my disks I run the command:
colmux -addr filename -command "-sD"
which essentially starts collectl running on all the machines listed in 'filename' and runs the -sD command which is collectl's moral equivalent of iostat -x, which reports all individual disks stats including queue depth and service times. only restriction is you need passwordless ssh to all the machines you want colmux talk to.
the real magic is the output of those commands are fed back to colmux who in turns dispays as many as will fit in a terminal window, sorting by the column of your choice. in other word, you could have over 1000 disks and colmux will report the n-slowest. makes it possible to find that needle-in-the-haystack with ease...
in other words, think 'top anything' across an entire cluster!
-mark
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
