memory usage slowly rising

Josh huber at paradoxical.net
Fri Feb 24 22:39:52 UTC 2006


Relevant information:

- linux 2.4.x under user mode linux
- guest has 64M of RAM allocated
- guest has 64M of swap allocated
- running tor vers 0.1.1.14 (official debian package on sarge)

I end up having to reboot my node about once every 3-4 days
because eventually all the VM in the guest gets allocated and the
kernel gets very angry.

I use:

BandwidthRate 200 KB
BandwidthBurst 250 KB

Also, I use the default exit policy.

I've been recording system memory usage over the past few days
and it just keeps steadily rising:

At Wed Feb 22 16:15:30 EST 2006:

        total:    used:    free:  shared: buffers:  cached:
Mem:  61734912 58466304  3268608        0  1880064 16465920
Swap: 67100672   290816 66809856

At Fri Feb 24 17:34:48 EST 2006:

        total:    used:    free:  shared: buffers:  cached:
Mem:  61734912 59625472  2109440        0   671744 29253632
Swap: 67100672 27660288 39440384

Notice the huge rise is swap used.

Currently, the tor process is consuming 70024 kB of virtual
address space, and has a 41296 kB RSS.

Unfortunately, I haven't been watching the tor process usage
until just now, but I will next time the uml guest is restarted.
Tor is effectively the only daemon running inside the
guest--privoxy and sshd are there, but are consuming only a
couple MB.

Anyone else running a tor node in UML?  The latest alphas have
been much better than the stable line in terms of CPU usage, but
they all seem to exhibit the same memory usage patterns.

I wonder if this is an artifact of running inside UML?

Josh



More information about the tor-talk mailing list