After rabbitmq failover neutron-server may get into state when it can't declare an exchange and consumes ~100% cpu
Bug #1493732 reported by
Eugene Nikanorov
This bug affects 1 person
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mirantis OpenStack |
Fix Released
|
Critical
|
Oleg Bondarev | ||
7.0.x |
Fix Released
|
Critical
|
Eugene Nikanorov | ||
8.0.x |
Fix Released
|
Critical
|
Oleg Bondarev |
Bug Description
Fuel build #287, scale lab 10
In certain conditions (high load, rabbitmq failover), neutron-server may get into state when it tries to declare an exchange in message queue brocker and fails.
That leads to a kind of busy-spin resulting in inability for agents to retrieve information from server via RPC.
Agents also can't report state to server which looks like some agents are down from time to time.
User impact: in such environment basic cloud features likely to work unstable or doesn't work at all (such as: spawning VMs, external network access, ability to receive fixed ip address for VM)
Changed in fuel: | |
status: | New → Confirmed |
no longer affects: | mos |
description: | updated |
Changed in mos: | |
assignee: | nobody → MOS Oslo (mos-oslo) |
status: | New → Confirmed |
importance: | Undecided → Critical |
Changed in mos: | |
milestone: | none → 7.0 |
no longer affects: | fuel |
no longer affects: | fuel/7.0.x |
no longer affects: | fuel/8.0.x |
tags: | added: scale |
tags: | added: 70mu1-confirmed |
tags: | removed: 70mu1-confirmed |
tags: | added: neutron |
tags: |
added: area-neutron removed: neutron |
tags: | added: 8.0 release-notes-done |
To post a comment you must log in.
This is a regression caused by commit during synchronising upstream Kilo puppet manifests to 7.0 https:/ /review. openstack. org/#/c/ 189678/ (parameter database_ max_pool_ size was set to 10) but even in 5.0 we set this parameter to 50 by https:/ /review. openstack. org/#/c/ 100859/ 4/deployment/ puppet/ openstack/ manifests/ controller. pp