openstack/kolla-ansible
Code Issues Proposed changes

[CI] Ironic: use ipmitool retries

Browse Source 
In Ironic jobs with Tenks, we saw issues with IPMI commands
failing, resuling in job failures:

  Error setting Chassis Boot Parameter 5

A metal3.io commit [1] was found that fixes the issue by moving IPMI
retries from ironic to ipmitool, which has a side-effect of increasing
the timeout. This change applies the same configuration.

This change has been adapted from an analogous change in
kayobe-config-dev. [2]

[1] 6bc1499d8b
[2] Ib4fce74cebebe85c31049eafe2eeb6b28dfab041

Co-Authored-By: Mark Goddard <mark@stackhpc.com>
Change-Id: I552417b9da03b8dfc9406e0ff644092579bc7122
This commit is contained in:
Radosław Piliszek 2022-03-07 10:20:08 +01:00
parent 91778aca76
commit a6145d678d
1 changed files with 22 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
tests/templates/ironic-overrides.j2
View File

@ -1,3 +1,25 @@
[neutron]
cleaning_network = public1
provisioning_network = public1
# This IPMI configuration has been taken from the metal3.io ironic-image
# project [1]. It is necessary to make IPMI reliable in Kolla Ansible Ironic
# jobs. Without it, we saw the following error:
# Error setting Chassis Boot Parameter 5
# [1] https://github.com/metal3-io/ironic-image/commit/6bc1499d8bb04c2c859b970b3739c3a8ed66ae2a
[ipmi]
# use_ipmitool_retries transfers the responsibility of retrying to ipmitool
# when supported. If set to false, then ipmitool is called as follows :
# $ipmitool -R 1 -N 1 ...
# and Ironic handles the retry loop.
use_ipmitool_retries = true
# The following parameters are the defaults in Ironic. They are used in the
# following way if use_ipmitool_retries is set to true:
# $ipmitool -R <X> -N <Y> ...
# where :
# X = command_retry_timeout / min_command_interval
# Y = min_command_interval
# If use_ipmitool_retries is false, then ironic retries X times, with an
# interval of Y in between each tries.
min_command_interval = 5
command_retry_timeout = 60