Hi,
We are on EL6 with CentOS 6.x. We want to improve the performance of ansible over SSH and we are hitting a roadblock with the accelerate mode.
We use multiple control nodes, so we are using the following configuration in the cfg file:
[accelerate]
accelerate_multi_key = yes
This however doesn’t seem to work. The test we are doing is this:
control node 1: run a playbook with accelerate=true. Playbook runs fine.
control node 2: run a playbook with accelerate=true. Playbook fails. with error: fatal: [xxxxxxxxx] => Failed to launch the accelerated daemon on xxxxxxxxxx (reason: failed to connect to the local socket file)
at this point of time if we go on the target node and kill the ansible socket, and then re-run the playbook from control node 2, then it runs fine. but it now fails from control node 1.
We then changed the configs on each control node so that they use different ports to connect:
on control node 1:
[accelerate]
accelerate_multi_key = yes
accelerate_port = 5099
on control node 2:
[accelerate]
accelerate_multi_key = yes
accelerate_port = 5098
However this still fails.
How can we make this work? Paramiko performance for us is extremely slow and this is causing a bottleneck for us.