Error: /Stage[main]/Role::Labs::Instance/Service[idmapd]: Could not evaluate: Could not find init script for 'idmapd'
on Debian Jessie instances
Description
Description
| Status | Subtype | Assigned | Task | ||
|---|---|---|---|---|---|
| Resolved | hashar | T103972 Setup jobs for operations/software/conftool | |||
| Resolved | hashar | T95545 Migrate all debian-glue jobs to Jessie slaves | |||
| Resolved | hashar | T102894 Move labs/tools/extdist tox tests to jessie | |||
| Resolved | hashar | T103977 Create tox templates that are bound to jessie | |||
| Resolved | hashar | T94836 Create CI slaves using Debian Jessie | |||
| Duplicate | None | T86728 Migrate Jenkins slaves from Ubuntu Trusty to Debian Jessie | |||
| Resolved | coren | T75592 Debian Jessie image for Labs | |||
| Resolved | coren | T87309 No init script for idmapd on labs jessie instances |
Event Timeline
Comment Actions
So, Debian ships an nfs-common package that contains a SysV init script with the same name that spawns all the needed daemons. Individual daemons like idmapd can be turned on/off from /etc/default/nfs-common. Ubuntu has forked the package and shipped upstart scripts that are separate for each daemon (because that way upstart can manage them better). Debian, even in jessie, does not ship systemd units for nfs-common at all, so the SysV init script is used instead.
This means that Debian doesn't have Service['idmapd'] but has Service['nfs-common'] instead. However, this also means that just handling the Service is not enough; /etc/default/nfs-common needs to also be puppet-managed.
All that said, I think @coren thinks that all this can go away when T87870 gets fixed. I'd probably bite the bullet and fix this independently, though, instead of waiting for the far larger fix to arrive.
Comment Actions
This bandaid has been merged about a month ago. Is there anything left here? (I think not)