quickstart should collect information on the BMC node and baremetal console
Bug #1785055 reported by
Gabriele Cerami
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
tripleo |
Fix Released
|
Critical
|
Gabriele Cerami |
Bug Description
When introspection or IPMI fail, we don't have much informations to work on for the debugging.
We should add to the log collection
- relevant IPMI logs from the BMC node
- Console transcript from the unprovisioned nodes.
Changed in tripleo: | |
milestone: | rocky-rc1 → rocky-rc2 |
Changed in tripleo: | |
milestone: | stein-1 → stein-2 |
Changed in tripleo: | |
milestone: | stein-2 → stein-3 |
Changed in tripleo: | |
status: | Triaged → Fix Released |
To post a comment you must log in.
I did some analysis of the requirements to get these additional information
Relevant IPMI logs from the BMC node:
We need to add the BMC node to the inventory, then add the relevant paths to the collect logs role for the BMC. To do this we need BMC IP and credentials to ssh.
The IP can be recovered in the instackenv.json pm_addr field.
The credentials are a bit more complex. In OVB the BMC node is created by the underlying cloud, using tenant credentials and a tenant generic private key. If we want to access it instead, we should use the undercloud key. So we need to change BMC creation slightly to use a different key. Difficult with te-broker, easier with direct provisioning.
Console transcript:
Here too, the situation is complex, overcloud nodes are created by the underlying cloud using ipxe images. To access their console, we need the tenant credentials.
Te broker has those and can recover the logs before deleting the node, but the difficult part is return these logs to the undercloud node for collection.
Direct provisioning again could make this a lot easier.