Searching an entire IPv6 subnet hangs the Machine Tracker

[Tim Chown]

We have dual-stack subnets/vlans in our network.

When accessing the Prefixes view from the Report area, we get a nice list of IPv4 and IPv6 prefixes in use, with a count of hosts in the right hand column (Active IPs).

If we click on a number for an IPv4 prefix, we see a list of the active IPs and associated MACs.

If we click on a number for an IPv6 prefix, the web interface hangs indefinitely.

The only factor we think might affect this is that we do not at this moment have reverse DNS configured for the prefix(es). We can certainly update the bug report when we fix this, but at present NAV isn't reporting the IPv6 IPs seen in a subnet from this screen.

[Morten Brekkevold (mbrekkevold)]

I've confirmed this on our development server. According to the logs, the machine tracker does actually retrieve some rows from the database, but it seems the rendering of the result causes the apache process to go into a tight loop and consume 100% CPU. I eventually gave up and restartet apache.

[Morten Brekkevold (mbrekkevold)]

To support the "inactive" checkbox, which causes the Machine Tracker to also output the inactive/unused addresses of a subnet, it will always iterate over all possible addresses in the range between the entered FROM and TO-addresses.

For a typical IPv4 /24 subnet this means 2^8 iterations, which is quite fast.

For a typical IPv6 /64 subnet, on the other hand, this means 2^64 iterations. Ouch. This is our tight loop right there.

I'm working on a fix to disable the inactive-function for IPv6 addresses.

[Morten Brekkevold (mbrekkevold)]

Fix committed to the 3.4.x series branch: https://metanav.uninett.no/hg/series/3.4.x/rev/9ea2376e78b0