)]}'
{"elements/puppet-stack-config/puppet-stack-config.pp":[{"author":{"_account_id":9555,"name":"Matthew Booth","email":"mbooth@redhat.com","username":"MatthewBooth"},"change_message_id":"ad3d1b6fb11c19f8b22760b3594bc5edb086490c","unresolved":false,"context_lines":[{"line_number":539,"context_line":"Keystone_endpoint \u003c||\u003e -\u003e Service[\u0027nova-compute\u0027]"},{"line_number":540,"context_line":"Keystone_service \u003c||\u003e -\u003e Service[\u0027nova-compute\u0027]"},{"line_number":541,"context_line":""},{"line_number":542,"context_line":"# This is a workaround for a race between nova-compute and ironic"},{"line_number":543,"context_line":"# conductor.  When https://bugs.launchpad.net/tripleo/+bug/1777608"},{"line_number":544,"context_line":"# is fixed this can be removed."},{"line_number":545,"context_line":"Service\u003c| title \u003d\u003d \u0027ironic-conductor\u0027|\u003e -\u003e Service[\u0027nova-compute\u0027]"},{"line_number":546,"context_line":""},{"line_number":547,"context_line":"if str2bool(hiera(\u0027enable_tempest\u0027, true)) {"}],"source_content_type":"text/x-puppet","patch_set":2,"id":"5f7c97a3_31a212fe","line":544,"range":{"start_line":542,"start_character":0,"end_line":544,"end_character":31},"updated":"2018-06-19 08:45:41.000000000","message":"Thanks :)","commit_id":"2e045e9b3f6ba76daa82e302afa348a6b0d748ff"},{"author":{"_account_id":9555,"name":"Matthew Booth","email":"mbooth@redhat.com","username":"MatthewBooth"},"change_message_id":"b7d5f3202861b189d3a5b33f76237f7f0ff509b2","unresolved":false,"context_lines":[{"line_number":548,"context_line":"  ensure  \u003d\u003e directory,"},{"line_number":549,"context_line":"  require \u003d\u003e Anchor[\u0027nova::install::end\u0027],"},{"line_number":550,"context_line":"}"},{"line_number":551,"context_line":"-\u003e file {\"/etc/systemd/system/${::nova::params::compute_service_name}.service.d/90-after-ironic-conductor.conf\":"},{"line_number":552,"context_line":"  content \u003d\u003e \"[Unit]\\nAfter\u003d${::ironic::params::conductor_service}.service\\n\","},{"line_number":553,"context_line":"}"},{"line_number":554,"context_line":"~\u003e exec { \u0027systemd daemon-reload for lp1777608\u0027:"},{"line_number":555,"context_line":"  command     \u003d\u003e \u0027systemctl daemon-reload\u0027,"},{"line_number":556,"context_line":"  path        \u003d\u003e [\u0027/usr/bin\u0027, \u0027/usr/sbin\u0027],"},{"line_number":557,"context_line":"  refreshonly \u003d\u003e true,"},{"line_number":558,"context_line":"  notify      \u003d\u003e Service[$::nova::params::compute_service_name]"},{"line_number":559,"context_line":"}"},{"line_number":560,"context_line":""},{"line_number":561,"context_line":""}],"source_content_type":"text/x-puppet","patch_set":4,"id":"5f7c97a3_5eb1b743","line":558,"range":{"start_line":551,"start_character":0,"end_line":558,"end_character":63},"updated":"2018-06-20 12:59:43.000000000","message":"To be clear, my -1 applies to this portion of the change. My reasoning is that we conceived this as a hack to workaround an issue, but because of the way service startup works, it doesn\u0027t workaround that issue. So even though this does exactly what we intended it to do, I\u0027m no longer in favour of it because it\u0027s just tech debt with no upside.\n\nTo be even clearer, we deliberately wanted this to be a very weak dependency between services. Specifically, we don\u0027t want a hard dependency between nova compute and ironic conductor which would mean, for eg, that you can\u0027t restart ironic conductor without also forcing a restart of nova compute. Consequently we went with just After\u003d.\n\nAfter\u003d only affects ordering *if the 2 services are already being explicitly started in the same transaction*, for example at boot time. So this only has any effect at all if we\u0027re already starting both things at the same time. What we discovered was that, because of the behaviour of oslo.service, systemd doesn\u0027t wait for initialisation to complete: the notification is essentially sent immediately. You observed a 10sec delay on a highly resource limited test system, in practise on any realistic system the delay will not be perceptible. In any case, even a 10sec delay doesn\u0027t buy us anything once we have the nova loop in place.\n\nSo for the same reason I wanted a comment on the puppet change (to aid the removal of a hack), I don\u0027t think we should add this hack in the first place.","commit_id":"2b3189f1a226c42f720686e67bd0cc3b2e4f8b17"}]}
