While upgrading the public transit relay, I noticed that the munin plugins were not reporting any of the traffic that occurred during the first few minutes of the process lifetime.
I realized that the per-connection started field is blurred (meaning rounded down to the next e.g. 5 minute period), but the rebooted field in the current table is not. The munin plugins calculate the "bytes transferred since process start" by filtering the usage rows on started > rebooted and summing them. But if the reboot happened just after a 5-minute window, most connections that complete during that window will be blurred into the past far enough to appear old.
This is at most a minor nuisance, but as we address the other papercuts in #4 we should think about this one too. One fix might be to change blur() to be aware of the reboot time, with something like max(reboot, rounded(start_time)). Another might be to blur the reboot time as well (although I don't think that's a good idea).
While upgrading the public transit relay, I noticed that the munin plugins were not reporting any of the traffic that occurred during the first few minutes of the process lifetime.
I realized that the per-connection
startedfield is blurred (meaning rounded down to the next e.g. 5 minute period), but therebootedfield in thecurrenttable is not. The munin plugins calculate the "bytes transferred since process start" by filtering the usage rows onstarted > rebootedand summing them. But if the reboot happened just after a 5-minute window, most connections that complete during that window will be blurred into the past far enough to appear old.This is at most a minor nuisance, but as we address the other papercuts in #4 we should think about this one too. One fix might be to change
blur()to be aware of the reboot time, with something likemax(reboot, rounded(start_time)). Another might be to blur thereboottime as well (although I don't think that's a good idea).