NOTICE: Downloading the consensus from maatuska took 31.4s. Median download time is 1.1s: moria1 => 1.0s, dizum => 0.4s, gabelmoo => 0.4s, dannenberg => 0.5s, maatuska => 31.4s, longclaw => 1.1s, bastet => 1.5s, faravahar => 2.0s NOTICE: The system clock of maatuska is 22 seconds off NOTICE: Authorities disagree about the BadExit flag for 7A38E62A305DD9CE704BF0F1FB6BF0D0F1BDC417 (with flag: moria1, gabelmoo, longclaw, without flag: faravahar)
gk@torproject.org wrote Wed, 9 Oct 2024 05:06:37 +0000 (UTC):
NOTICE: Downloading the consensus from maatuska took 31.4s. Median download time is 1.1s: moria1 => 1.0s, dizum => 0.4s, gabelmoo => 0.4s, dannenberg => 0.5s, maatuska => 31.4s, longclaw => 1.1s, bastet => 1.5s, faravahar => 2.0s NOTICE: The system clock of maatuska is 22 seconds off
maatuska's clock doesn't seem to be or have been off recently, judging from system logs.
From reading doctor/consensus_health_checker.py it seems like
clock_skew, on which the warning about system clock being off is based, assumes that the time it takes between request being sent and authority composing a reply header is zero.
If that's true it follows that with a correct system clock on the authority, the warning indicates how much time was spent between doctor sending the request and the authority handling it.
On Wed, Oct 09, 2024 at 02:55:16PM +0200, Linus Nordberg via tor-consensus-health wrote:
From reading doctor/consensus_health_checker.py it seems like clock_skew, on which the warning about system clock being off is based, assumes that the time it takes between request being sent and authority composing a reply header is zero.
If that's true it follows that with a correct system clock on the authority, the warning indicates how much time was spent between doctor sending the request and the authority handling it.
Agreed!
I made a ticket about this bug: https://gitlab.torproject.org/tpo/network-health/doctor/-/issues/40017
--Roger
tor-consensus-health@lists.torproject.org