Opened 7 months ago

Last modified 7 months ago

#32975 new defect

Tor Socks unresponsive behaviour after 24-36 hours of runtime

Reported by: ntrn Owned by:
Priority: Medium Milestone:
Component: Core Tor/Tor Version: Tor: 0.4.2.5
Severity: Normal Keywords:
Cc: Actual Points:
Parent ID: Points:
Reviewer: Sponsor:

Description

Error description:

After around 1 or 1.5 days the tor socks proxy is getting unresponsive. There is actually no error message from tor itself. Only nyx is claiming to not have available buffer (error 105) and does not start any more.

Error message:

nyx -c .nyx/config -i 9951

Unable to connect to 127.0.0.1:9951: [Errno 105] No buffer space available

Workaround:

I could solve the problem by deleting the ~/.tor folder and restarting tor.

Environment:

  • 2 CPU VPS with 4GB RAM
  • Tor version 0.4.2.5 build and installed from source
    (also seen this behaviour on CentOS 7 repo package of tor v0.3.5.8)
  • CentOS 7
    (Linux version 3.10.0-042stab139.1 (root@…) (gcc version 4.4.7 20120313 (Red Hat 4.4.7-18) (GCC) ) #1 SMP Tue Jun 18 12:51:14 MSK 2019)
  • non-exit relay
  • controle port blocked by iptables for outside connection
  • ssh port blocked by iptables for outside connection

Child Tickets

Change History (3)

comment:1 Changed 7 months ago by ntrn

activated "ConstrainedSockets 1" in torrc.
Testing behaviour.

comment:2 Changed 7 months ago by ntrn

did not solved the issue.
Have to reboot the server at least daily to let it run stable.

comment:3 Changed 7 months ago by arma

Can you point us to which relay this is?

What are you doing to decide that the socksport is unresponsive? Or is it something else, since you say 'socks proxy' not socksport?

The "No buffer space available" error seems weird and like a good hint: maybe the vps is artificially constrained in some resource like sockets or kernel buffers. I'd check 'dmesg' too.

Note: See TracTickets for help on using tickets.