Do you have something like strace in NetBSD to measure the syscall times on the backend filesystem? It will be helpful to eliminate the backend variable (by measuring time taken) during the slow runs..
Avati
On Wed, May 8, 2013 at 5:15 PM, Emmanuel Dreyfus <manu@xxxxxxxxxx> wrote:
Anand Avati <anand.avati@xxxxxxxxx> wrote:Note that the performance issue survives restarting the daemons,
> The " Undefined error: 0."
> certainly looks new and suspicious.
therefore I can patch to test things.
