[tahoe-dev] Possible memory leak?

[Michael Coppola]

Hi list,

I am not sure if this is a bug or simply my improper usage of Tahoe, but
it seems that after uploading a large file or large number of files, the
Tahoe daemon allocates a huge amount of RAM and doesn't unallocate it.

To explain it a little further, I have 1gb of RAM on one of my storage
nodes.  A fresh restart of the daemon will leave me with 723mb of RAM
free.  I upload three 60mb files at the same time over FTP (using this
node as the gateway), and RAM usage progressively increases until I am
left with 407mb RAM free.  This number holds steady.  A few minutes
later, I upload three 170mb files, and the amount of free RAM hovers
around 15mb, eventually down to 1 or 2.

At this point, the OS begins swapping and performance deteriorates until
subsequent uploads completely timeout, and it hangs just trying to
create a new directory.  I restart the daemon (`tahoe restart .`), and
everything is fine again.

I'm running CentOS 5.5 using Tahoe 1.8.2.  Any ideas?

Thank you,
Michael