rabbitmqctl list commands are not resilient to network delays
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mirantis OpenStack |
Invalid
|
Medium
|
Alexey Lebedeff | ||
10.0.x |
Won't Fix
|
Medium
|
Alexey Lebedeff | ||
9.x |
Won't Fix
|
Medium
|
Alexey Lebedeff |
Bug Description
Detailed bug description:
list_channels and list_queues are used by OCF script for monitoring, but this commands can fail in not-so-perfect network conditions. There is an upstream improvement that removes rpc roundtrip and makes this commands mostly ignore network delay - https:/
Steps to reproduce:
Introduce some small network delay using `tc` (e.g. 1ms) and look at how time it takes to list significant amount of queues skyrocket.
Expected results:
Listing big number of queues/channels should always take approximately some amount of time, irregarding of network delay - examples of acceptable and unacceptable values are at https:/
Actual result:
See comparsion of old and new implementations - https:/
Reproducibility:
Always
Workaround:
Use faster/less congested network.
Impact:
False positives in OCF script could result in unnecessary rabbitmq restarts.
Changed in mos: | |
status: | New → In Progress |
importance: | Undecided → High |
importance: | High → Medium |
assignee: | nobody → Alexey Lebedeff (alebedev-a) |
milestone: | none → 9.0 |
Changed in mos: | |
status: | In Progress → Won't Fix |
Changed in mos: | |
status: | Won't Fix → Confirmed |
tags: | added: 10.0-reviewed |
Fix proposed to branch: master /review. fuel-infra. org/19186
Change author: Alexey Lebedeff <email address hidden>
Review: https:/