[Swan-dev] [Testing] Test Suite & Docker

Andrew Cagney andrew.cagney at gmail.com
Thu May 12 15:01:57 UTC 2016


Here's a brain dump:

- yes 9p is isn't reliable, and it seems to be getting worse; I really
wonder about eliminating 9p and using copy/rsync/clone instead, this
might fit better with docker

- f22 also had a nice bug where cloning corrupted the root filesystem

- I've found trying to use more than #cores/2 (1 core for each KVM,
one for userland 9p?) doesn't seem to make things faster
  Docker could help, however ...

- "half" of each test is spent twiddling thumbs waiting for things to
timeout, docker won't help with that

- otoh, the other "half" of each test is spent waiting for _Fedora_ to
boot; given other OSs and linux variants boot in seconds there's
something going really really wrong
  Deploying docker instances containing rebuilt pluto should be faster
and more robust?  I've found cloning test domains to be really fast,
but getting the fresh clone to boot to be really slow; Fedora again.

- I'm guessing that, like the current KVM tests, docker doesn't do
FIPS "correctly" (as in run an entire FIPS stack including the
kernel); for the KVMs it is a small matter of programming, for docker?
 This might mean keeping both around

- Similarly, if we want to test against systems that are not amd64,
KVM would be need

Andrew


More information about the Swan-dev mailing list