Cluster gets stuck in "waiting for CNI"
Bug #1841259 reported by
Cory Johns
This bug affects 1 person
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| Flannel Charm |
Fix Released
|
Medium
|
Cory Johns | ||
Bug Description
The kubernetes-cni interface only watches for the changed flag to manage its other flags and not joined. There is a rare situation that can occur where the changed flag gets set before the joined flag (probably an upstream bug in charms.reactive). If this happens and the broken handler is run after the changed handler, then all of the flags get cleared out and the changed handler will never be called again, leading to the cluster being stuck waiting for CNI, while the CNI charm is stuck waiting for the flags from the cluster.
The easiest fix is to have the interface layer watch for either joined or changed, since even if the changed flag doesn't get set again, the joined one eventually does which will restore the flags.
Revision history for this message
| Cory Johns (johnsca) wrote | #1 |
| Changed in charm-flannel: | |
| status: | New → In Progress |
| assignee: | nobody → Cory Johns (johnsca) |
| Changed in charm-flannel: | |
| status: | In Progress → Fix Committed |
| milestone: | none → 1.16 |
| importance: | Undecided → Medium |
| Changed in charm-flannel: | |
| status: | Fix Committed → Fix Released |
To post a comment you must log in.
https:/