Tor Exit Relay CPU Usage Running at 100% for 1 MB/s on FreeBSD
Hi tor-relays@ mailing list, I have set up two exit relays on a FreeBSD 12.0 dedicated server: https://metrics.torproject.org/rs.html#details/1CD029594B08E07F29B9420410C2E... https://metrics.torproject.org/rs.html#details/A0EB2BD840838FAD51BDAD86B0BA3... Looking at my top stats, I get CPU usage of 100% most of the time (meaning 95% of the time) on both instances pushing around ~1 MB/s with both instances. The server is a HP Blade server and has a single Intel Xeon L5520 CPU with 16GB of RAM. Tor is configured to have 150 Mbps per instance, with a total of two instances. Why am I getting this abnormal CPU usage? What's the solution? I don't want to use Linux but can use another *BSD or Solaris/Illumos if I must. -Neel === https://www.neelc.org
On 3/11/19, Neel Chauhan <neel@neelc.org> wrote:
I have set up two exit relays on a FreeBSD 12.0 dedicated server: Looking at my top stats, I get CPU usage of 100% most of the time (meaning 95% of the time) on both instances pushing around ~1 MB/s with both instances.
HP Blade single Intel Xeon L5520 CPU with 16GB of RAM.
4 core HT 2.26GHz 32k/256k/8M-shared DDR3-800 x64 60W Nehalem from 2009 without AES... a 1st gen i7. Not having AES will impact things but said 8-16Mbps there should be trivial unless you're also busy crunching tons of key exchanges, connections, circuits. Running x32 lernel on x64 hw isn't good. Busy ZFS on x32 can be slow but tor doesn't abuse disk. Check power / cpu / thermal management on blades too. Tor from 'latest' pkg repo should be fine. Other than that, not enough info. You're going to have to play around with top's thread / io / cpu modes (hit '?'), netstat, vmstat, systat, gstat, sockstat, ps, etc... see whats going on.
participants (2)
-
grarpamp -
Neel Chauhan