Accelerate mode failing with multiple control nodes on CentOS 6.x

[Sunny Saxena]

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.

[Michael DeHaan]

what version of Ansible are you using?

 a fix for accelerate mode is applied in 1.7.2

--
You received this message because you are subscribed to the Google Groups "Ansible Project" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ansible-proje...@googlegroups.com.
To post to this group, send email to ansible...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/ansible-project/4be46051-35e3-4cd3-aebc-a3be53cd6a5d%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.