Weird problem with network reuse of TCP ports from apps started from ansible

Archives Ansible Project
1

So we have been running into an issue in our ansible rollout that hasn’t occurred before sometime in mid April. About that time we started getting conflicts with services reusing the 5099 fireball port if those processes were started from ansible.

In the latest case, I set up a new hardware box and ran the ansible jobs which load up various services like iscsid, nrpe, etc. All of these usually pull a port from the 10000->65536 when started by themselves, but everyone of them pulled 5099 as their listening port. I had to end up stopping about a dozen processes before I could run ansible again.

What I am wanting to know is if this is known or seen anywhere else and if there is some flag we might have setup which is triggering this that we could turn off.

Thanks.

2

Go ahead and open a github issue for this, and we’ll see if we can reproduce it.

Thanks!

3

We suspect this may be something needing to close file descriptors.

(BTW, just curious, you said fireball, are you using the new accelerate mode or still calling fireball directly? I remember you guys had some EL hosts so couldn’t do -c ssh yet since no Control Persist support just yet, though that + pipelining are highly recommended for those that can run them.

4

I call it fireball because that was what it was called back in my day. You
people and your new ways to make things go faster. Get off my sh lawn dag
nabbit

Sorry you are right I meant accelerate. I have applied a one line patch
from your team and am testing to see if I can get the conflict.