[tor-relays] 100K circuit request per minute for hours killed my relay

[Roman Mamedov]

On Fri, 21 Jul 2017 10:24:39 +0100
Pascal Terjan <pterjan at gmail.com> wrote:

> Last night for 4h30 (until VPS provider shut it down) one of my middle
> relays seems to have got in a bad state where it was using 100% CPU
> continuously
> 
> It was not using much bandwidth, about 4MB/s, but reading the logs it
> seems it was getting a lot of circuit requests (70K/minute initially,
> 100K at the end). Is there anything I can do about it?

Your message prompted me to check logs, and on one relay I see the following:

Jul 19 03:26:03.000 [notice] Circuit handshake stats since last time: 4242/4242 TAP, 65583/65583 NTor.
Jul 19 09:26:03.000 [notice] Circuit handshake stats since last time: 4897/4897 TAP, 60996/60996 NTor.
Jul 19 15:26:03.000 [notice] Circuit handshake stats since last time: 3682/3682 TAP, 1759242/1759519 NTor.
Jul 19 21:26:04.000 [notice] Circuit handshake stats since last time: 3762/3762 TAP, 5946906/5947506 NTor.
Jul 20 03:26:04.000 [notice] Circuit handshake stats since last time: 1676/1676 TAP, 6794460/6795903 NTor.
Jul 20 09:26:04.000 [notice] Circuit handshake stats since last time: 4432/4433 TAP, 6991546/6993275 NTor.
Jul 20 15:26:04.000 [notice] Circuit handshake stats since last time: 7649/7650 TAP, 1437786/1437748 NTor.
Jul 20 21:26:04.000 [notice] Circuit handshake stats since last time: 7484/7484 TAP, 68388/68388 NTor.
Jul 21 03:26:04.000 [notice] Circuit handshake stats since last time: 6105/6105 TAP, 52027/52027 NTor.
Jul 21 09:26:04.000 [notice] Circuit handshake stats since last time: 5210/5210 TAP, 99365/99365 NTor.

That one has not been shut down (the VPS provider does not seem to have any
issue with allowing high CPU usage) and works normally after that.

-- 
With respect,
Roman