5

I'm monitoring a VM functioning as our production server. I noticed it has at many times, quite low entropy (/proc/sys/kernel/random/entropy_avail).

On average our server(which among other stuff does SSL, session-key generation etc) runs with about 100-200 bits of available entropy.

Is this anything to be concerned with?

Henrik Kjus Alstad
  • 153
  • 5

1 Answers1

3

That sounds normal for a vm. You don't typically worry unless available entropy drops below that 100-200 average. The key measure is the sequences that come out of that entropy.

You should use the ent program to test your random sequences. You worry when you start seeing tests fail here.

This site has a good process for testing randomness.

We can collect random data from /dev/random using "dd". This command will collect 8,192 blocks from /dev/random and output the results to the text file we will call "random_output". NOTE: if your /dev/random generator is too slow this command will take forever to finish. The random_output file should be around 1 to 4 megabytes in size depending on how large a chuck is on your system. For now, if the dd command does not finish in 3 seconds or so hit Ctrl-c to kill dd.

dd if=/dev/random of=random_output count=8192

Then, after you've built your ent program:

./ent random_output

Analyzing the results is a little beyond the scope of your question, but what I linked can fill in the ancillary details.

Community
  • 1
avgvstvs
  • 940
  • 1
  • 7
  • 19
  • 1
    Awesome suggestion :) thanks. Just adding that if you're running a debian-based system, chances are that the ent program is very likely available through the package manager – Henrik Kjus Alstad Apr 27 '15 at 11:26