Uniter state file empty after node crash
Bug #1525868 reported by
Gabriel Samfira
This bug affects 1 person
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| Canonical Juju |
Won't Fix
|
High
|
Unassigned | ||
Bug Description
After a node crashed and came back up, the uniter state file for one of the units deployed was empty. As a consequence, the unit agent could not recover. Relevant abstract from logs:
http://
Contents of uniter file:
http://
Juju version 1.25.1, fresh bootstrap.
Revision history for this message
| Gabriel Samfira (gabriel-samfira) wrote | #1 |
Revision history for this message
| Gabriel Samfira (gabriel-samfira) wrote | #2 |
Revision history for this message
| Gabriel Samfira (gabriel-samfira) wrote | #3 |
| Changed in juju-core: | |
| status: | New → Triaged |
| importance: | Undecided → High |
Revision history for this message
| Gabriel Samfira (gabriel-samfira) wrote | #5 |
| Changed in juju: | |
| status: | New → Triaged |
| importance: | Undecided → High |
| milestone: | none → 2.1.0 |
| no longer affects: | juju-core |
Revision history for this message
| Anastasia (anastasia-macmood) wrote | #6 |
| Changed in juju: | |
| milestone: | 2.1.0 → none |
Revision history for this message
| Heather Lanigan (hmlanigan) wrote | #7 |
| Changed in juju: | |
| status: | Triaged → Won't Fix |
To post a comment you must log in.
A personal suggestion on this.
There is no way to prevent files from going away in some cases. Node crashing, accidental file removal, whatever. However in cases such as this, I think its acceptable if the state file is invalid, that it should be replaced with a boilerplate default state file. At worst, this means that all the hooks will be run again (which should be idempotent anyway), which is better then never being able to recover at all.