Specifically, innodb status samples, and vmstat, are not proportional to that. Instead, you get samples that are separated at about sleep_time (so for a sleep time of 60, you get two innodb status captures separated 60 seconds).
I think that somehow should be taken into account. We can either submit a patch to pt-stalk to handle all the data capture processes the same, or see if using large values for sleep time makes sense for these specific captures (I think it does not).
This is a question, not a bug report, btw.
Specifically, innodb status samples, and vmstat, are not proportional to that. Instead, you get samples that are separated at about sleep_time (so for a sleep time of 60, you get two innodb status captures separated 60 seconds).
I think that somehow should be taken into account. We can either submit a patch to pt-stalk to handle all the data capture processes the same, or see if using large values for sleep time makes sense for these specific captures (I think it does not).
This is a question, not a bug report, btw.