[tor-relays] TOR Relay performance issue

[Thomas Maurice]

Hello all ! o/

It seems I am running into a performance issue with my relay, and I am
not able to troubleshoot it, so I hope someone could point out something
obvious I am missing!

(for information, the relay is a 2vCPU, 2G RAM VM running on Scaleway)

From what I observe from the graphs, my maximal CPU usage ceils at ~60%,
and the load ceils at 1 eventhough the TOR daemon is aware I have 2
cores available, and I set the NumCPUs configuration option accordingly)

First I tought that the network did not want to send me more traffic,
but my graphs are way too regular to make this eventuality plausible, so
I am pretty convinced that it is due to a lack of optimization on my side.

(link to my graphs for reference
http://www.noelshack.com/2017-01-1483803911-constancesurvingt.png this
is a 1 hour snapshot, but it has been this constant for at least one day)

My first hypothesis was that the daemon is running out of file
descriptors, so I raised the limit that was in the defaults (doubled it
actually, by editing /etc/init.d/tor like a pig, to test), but I did not
observe any change in Grafana.

It is supposed to be caped at 20MBps, and 25MBps burst so I doubt the
limitation I am observing is enforced by the node itself.

If anyone had any pointers I would be grateful!

Cheers!
-- 
Thomas

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 858 bytes
Desc: OpenPGP digital signature
URL: <http://lists.torproject.org/pipermail/tor-relays/attachments/20170107/5df8aa1a/attachment.sig>