[RFE] Allow to not cache images on conductor for iPXE nodes in standalone mode
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Ironic |
Won't Fix
|
Wishlist
|
Pavlo Shchelokovskyy |
Bug Description
In integrated mode, Ironic already supports skipping caching of deploy image kernel/ramdisk on the conductor node when iPXE is used (through pxe.ipxe_use_swift option) and not serving those images from per-conductor HTTP servers, in which case node boots into deploy image directly from Swift TempURLs.
It is proposed to add support for similar functionality for standalone Ironic.
As our currently preferred deploy method with agent already requires that user images are stored on a HTTP server available from inside provisioning network, iPXE-capable nodes can boot to deploy image if this is stored on the same server as well and not cache them locally.
For this it is proposed to add a config option pxe.ipxe_
Alternatively, to support a more heterogeneous set of nodes, instead of global config option a new node capability field of the same name (also, defaulting to an empty list for backward compatibility) can be introduced and checked against when preparing the node boot to deploy image.
This can simplify ironic-conductor deployment by not hard requiring the per-conductor HTTP server, improve performance for distributed standalone Ironic as the (static) external HTTP server can be scaled horizontally, and make the ironic-conductor service more state-less.
Changed in ironic: | |
assignee: | nobody → Pavlo Shchelokovskyy (pshchelo) |
summary: |
- Allow to not cache images on conductor for iPXE nodes in standalone mode + [RFE] Allow to not cache images on conductor for iPXE nodes in + standalone mode |
Changed in ironic: | |
importance: | Undecided → Wishlist |
Changed in ironic: | |
status: | Triaged → Won't Fix |
The patch seems to be here, not sure why the bug # is different - https:/ /review. openstack. org/367321 seems pretty straight forward to me