sabnzbdplus-0.6.8-1 HTTPS CLOSE_WAIT problem

[blammo]

(starting here, since I'm running 0.6.8-1 on ubuntu 3.0.0-15-server)

fresh install of Ubuntu 11.10, patched, upgraded, installed sabnzbdplus, running behind a pfsense firewall, monitoring HTTPS with Chrome/Firefox "sendto" plugins, like sabconnect and sabplus.

over a fairly short period of time (4-12 hours), the sabnzbdplus process will stop responding on the HTTPS port (verified by telnet localhost PORT, which will open, but just sits non-responsive) even though the process itself is still running, downloading,etc. Looking at the netstat shows a lot of ports in "CLOSE_WAIT", the more ports the longer I wait to restart it. It's hung as we speak, with a fairly low count:

21 ESTABLISHED
38 CLOSE_WAIT

the error log shows complains about "ssl handshake failure"

==> /var/log/sabnzbd.error.log <==
File "/usr/share/sabnzbdplus/cherrypy/wsgiserver/__init__.py", line 1078, in _safe_call
return call(*args, **kwargs)
File "/usr/share/sabnzbdplus/cherrypy/wsgiserver/__init__.py", line 748, in sendall
bytes_sent = self.send(data)
File "/usr/share/sabnzbdplus/cherrypy/wsgiserver/__init__.py", line 1129, in send
return self._safe_call(False, super(SSL_fileobject, self).send, *args, **kwargs)
File "/usr/share/sabnzbdplus/cherrypy/wsgiserver/__init__.py", line 1108, in _safe_call
raise FatalSSLAlert(*e.args)
FatalSSLAlert: [('SSL routines', 'SSL23_WRITE', 'ssl handshake failure')]

It's not an open files problem, found and fixed that one early.

open files (-n) 32768

Restarting the process always cures the issue, but it's frustrating. Yesterday I worked through every chrome/IE/FF plugin I could find to monitor, and the problem is still happening.

Ideas?


(edit: just realized I wasn't running the latest from the "unofficial" repo (ppa) - going to give that a shot)

[sander]

Yes, upgrading to the newest version (0.6.14) is always a good start.

This problem has been discussed a few months ago on this forum. From my memory: it happened with browser plugins activated (not caused with normal browser HTTPS traffic). I think disabling the plugin solved / worked-around it. The problem was also reported 'upstream', but I can't remember whether the upstream was the cherrypy / SSL framework, or in the plugin.
I believe there was no clear solution; the thread or the problem just faded ... :-(

[sander]

Ah, found the threads:

http://forums.sabnzbd.org/viewtopic.php ... 89&p=53449

http://forums.sabnzbd.org/viewtopic.php?t=8644

Plugin was sabconnect++ , see http://code.google.com/p/sabconnectplus ... tail?id=94 , however see closing remark

[blammo]

12+ hours and running without issue.. upgraded to 0.6.14, and took recommendations from other threads to disable plush. Together, problem is hiding, hopefully gone for good.

[sander]

12+ hours and running without issue.. upgraded to 0.6.14, and took recommendations from other threads to disable plush. Together, problem is hiding, hopefully gone for good.

Did you 'only' upgrade to 0.6.14, and are you still running sabconnect++ ? Or did you stop using sabconnect++ ?