WIP: Allow for control-plane node replacement in podman-etcd #2061
Conversation
…ties of podman-etcd to allow for control-plane node replacement.
|
Can one of the admins check and authorise this run please: https://ci.kronosnet.org/job/resource-agents/job/resource-agents-pipeline/job/PR-2061/1/input |
| return "$OCF_ERR_GENERIC" | ||
| fi | ||
|
|
||
| # When only the local revision exists, it can start normally. This ensures that it can start during node replacement events. |
There was a problem hiding this comment.
I think this is too risky, as it be part of any normal startup, not only when an admin operates a node replacement.
There is an alternative already, the admin can set the property "force_new_cluster" on the node will ignore the revisions.
| return "$OCF_SUCCESS" | ||
| fi | ||
|
|
||
| # When only the peer revision exists, the local node must join as a learner. This ensures that it can start during node replacement events. |
There was a problem hiding this comment.
This is less risky, as the local revision comes directly from disk, so this might occur only if the node has no etcd data at all.
| ocf_log err "local revision is older and peer is not starting: cannot start" | ||
| ocf_exit_reason "local revision is older and peer is not starting: cannot start" | ||
| ocf_log err "local revision is older or empty and peer is not starting: cannot start" | ||
| ocf_exit_reason "local revision is older or emptyand peer is not starting: cannot start" |
There was a problem hiding this comment.
Missing a space between "and"
| else | ||
| if [ "$(attribute_node_cluster_id get)" = "$(attribute_node_cluster_id_peer)" ]; then | ||
| ocf_log info "same cluster_id and revision: start normal" | ||
| else if [ "$revision_compare_result" = "only_local" ] && [ "$(attribute_node_cluster_id get)" != ""] && [ "$(attribute_node_cluster_id get)" != "null" ]; then |
There was a problem hiding this comment.
A couple of minor notes on this
- you can use elif here
- there's a missing space before a closing bracket "]"
Moreover, this seems wrong to me. We have 2 starting resources with different cluster-ids, they cannot start normally.
This is a collection of tweaks intended to adjust the start-up properties of podman-etcd to allow for control-plane node replacement.