Windows 10 tor relay use 100% CPU

Question

I setup tor relay on Windows 10 see https://metrics.torproject.org/rs.html#details/AD5B534F1CF8ABD17001817CC2ACF6D11345766F as windows service.

At service start it use about 15% CPU. Then when it works correct it can use from 0 to 5 % CPU.

But some times tor service start to use 100% CPU for a long time (10 and more minutes). So I restart tor service. It can happened 2 or times per day.

And at this time it does not send or receive anythings by network and it does not read or write anything to disk. Network and disk usage are 0%.

Why tor service utilize 100% CPU? Is it bug on Windows or I have configuration errors?

---

I see next log messages when tor service use 100% CPU

Feb 24 00:15:54.000 [info] store_multiple(): Adding consensus diff from BDCF9DAE9B06BE846F509A2CD4622D0296F4C524A5B97FD55B87A19082BFFD44 to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with identity
Feb 24 00:15:54.000 [info] store_multiple(): Adding consensus diff from BDCF9DAE9B06BE846F509A2CD4622D0296F4C524A5B97FD55B87A19082BFFD44 to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with gzip
Feb 24 00:15:54.000 [info] store_multiple(): Adding consensus diff from A2C8550157121D20120939A0CE806E8C80F9EAD261EEF569840D433EC4AADA83 to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with identity
Feb 24 00:15:54.000 [info] store_multiple(): Adding consensus diff from A2C8550157121D20120939A0CE806E8C80F9EAD261EEF569840D433EC4AADA83 to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with gzip
Feb 24 00:15:54.000 [info] store_multiple(): Adding consensus diff from 96CCFFE34D19FE0E9A271C171C9E29EE2D95A309CBCDA05071F65B9BDE0F05AC to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with identity
Feb 24 00:15:54.000 [info] store_multiple(): Adding consensus diff from 96CCFFE34D19FE0E9A271C171C9E29EE2D95A309CBCDA05071F65B9BDE0F05AC to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with gzip
Feb 24 00:15:54.000 [info] store_multiple(): Adding consensus diff from 743550945E24E932FA2A4548160FB7B94A56E8DB36F0024BDBB0DA0B0C084CBB to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with identity
Feb 24 00:15:54.000 [info] store_multiple(): Adding consensus diff from 743550945E24E932FA2A4548160FB7B94A56E8DB36F0024BDBB0DA0B0C084CBB to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with gzip
Feb 24 00:15:55.000 [info] store_multiple(): Adding consensus diff from 6E83B5ED1443A63C1D2A227E40C14876CE9A444D31F2797EBEBD3131875F9804 to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with identity
Feb 24 00:15:55.000 [info] store_multiple(): Adding consensus diff from 6E83B5ED1443A63C1D2A227E40C14876CE9A444D31F2797EBEBD3131875F9804 to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with gzip
Feb 24 00:15:55.000 [info] store_multiple(): Adding consensus diff from C1733C9DB3D3FFA41648B4512D3ED9088706828C0B46B63910F9334CD18D71B3 to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with identity
Feb 24 00:15:55.000 [info] store_multiple(): Adding consensus diff from C1733C9DB3D3FFA41648B4512D3ED9088706828C0B46B63910F9334CD18D71B3 to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with gzip
Feb 24 00:15:55.000 [info] store_multiple(): Adding consensus diff from D21909A5CCEAE66D820DCB25D5BBF6591B8E4D7199FBE52C6F2B4D7607F445D6 to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with identity
Feb 24 00:15:55.000 [info] store_multiple(): Adding consensus diff from D21909A5CCEAE66D820DCB25D5BBF6591B8E4D7199FBE52C6F2B4D7607F445D6 to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with gzip
Feb 24 00:15:55.000 [info] store_multiple(): Adding consensus diff from 244325664A3D3FEA6EEE708E3A7F5457B9313C108C744275A7A6BAB031E1783F to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with identity
Feb 24 00:15:55.000 [info] store_multiple(): Adding consensus diff from 244325664A3D3FEA6EEE708E3A7F5457B9313C108C744275A7A6BAB031E1783F to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with gzip
Feb 24 00:15:55.000 [info] store_multiple(): Adding consensus diff from 8A68AF897614AB690785BF1C709249ADCDEA2E78657B41CE55285FAE3D3672CE to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with identity
Feb 24 00:15:55.000 [info] store_multiple(): Adding consensus diff from 8A68AF897614AB690785BF1C709249ADCDEA2E78657B41CE55285FAE3D3672CE to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with gzip
Feb 24 00:15:56.000 [info] store_multiple(): Adding consensus diff from 1520D1CA72D0806C7F627C57ED9B1380904CFD24F504762A4B4649DF1B70CC4B to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with identity
Feb 24 00:15:56.000 [info] store_multiple(): Adding consensus diff from 1520D1CA72D0806C7F627C57ED9B1380904CFD24F504762A4B4649DF1B70CC4B to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with gzip
Feb 24 00:15:56.000 [info] store_multiple(): Adding consensus diff from ACFCFDD29853CA1EB252867E1643F9B7A2C389DD2BC5B4F3FC44EBA8F2D772CC to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with identity
Feb 24 00:15:56.000 [info] store_multiple(): Adding consensus diff from ACFCFDD29853CA1EB252867E1643F9B7A2C389DD2BC5B4F3FC44EBA8F2D772CC to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with gzip
Feb 24 00:15:57.000 [info] store_multiple(): Adding consensus diff from 09783C82602523E03D46483573F927E768EC2C70E1008956EEEFFA80185EC0AB to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with identity
Feb 24 00:15:57.000 [info] store_multiple(): Adding consensus diff from 09783C82602523E03D46483573F927E768EC2C70E1008956EEEFFA80185EC0AB to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with gzip
Feb 24 00:15:57.000 [info] store_multiple(): Adding consensus diff from 39AE8F1436B0C78B5A39A90D46EB46A960E30253595DB9B6EC74E80B8532185D to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with identity
Feb 24 00:15:57.000 [info] store_multiple(): Adding consensus diff from 39AE8F1436B0C78B5A39A90D46EB46A960E30253595DB9B6EC74E80B8532185D to C195F4A134922B69D8DCA02C1C9D92586102A99, compressed with gzip

And so on...

Answer 1

Pretty sure its a knows issue. I had same problem and through googling i found you can set its process priority to below normal and its affinity to only use one core in task manager - details - right click on tor.exe. That way when it maxes out, its not top priority over other things, and it can only max out one core. Dosent seem to effect the relays performance at all.

Answer 2

This is ticket for problem. It is caused by diff cache, there is also workaround.

https://trac.torproject.org/projects/tor/ticket/24857

Answer 3

Personally, I don't recommend running a relay on Windows. It is well known to have many issues, and extremely buggy. Most relays run on Linux systems because they tend to be more reliable, and Windows systems tend to be more vulnerable because malware is engineered for the OS

In fact, I don't believe there is much if any official documentation for Windows.

If you insist on running on Windows, you will run into many bugs and performance issues. I personally run a relay on ubuntu server, and only see roughly 29% usage of CPU (dual core, 2.1 GHz) and ~400MB RAM

One other possibility is your CPU just can't handle the load. Though, any modern CPU of 2GHz plus should do