Date |
|
Authors | |
Status | In-ProgressDocumentation |
Summary | |
Impact | Applications were unable to fully use the message broker until RabbitMQ stopped synchronising nodes. After that a number of processing changes went through in significant numbers including overnight updates for Revalidation. |
...
Non-technical Description
...
Trigger
.Automated maintenance with large queue
...
Detection
.Alerts in monitoring channel
...
Resolution
Stopped the synchronisation across nodes in the cluster
Horizontal scaling consumers of queues with lots of message
...
Timeline
22:00-23:59 - Message Broker maintenance window
22:50 - Broker nodes begin being restarted
22:56 - One of the secondary nodes becomes unreachable from the primary
23:28 - Primary and 1 secondary enter an “Memory Alarm”state, although memory reported for the broker did not reflect this
08:41 - Stopped syncing a third node for of 2 queues being synchronised First of 3 failed attempts of a trainee to digitally sign conditions of joining
08:45 - Picked up alert and began identifying the issue. We paused some of the feeds in order to free up some
- Stopped the sync process for a (new) node for 1 of 2 queues that was syncing
12:11 - Cancelled Sync of largest queue
12:13 - Other queue finished synchronising
- Further maintenance window forced synchronisation, this time with much smaller queues
5 Whys (or other analysis of Root Cause)
We received “connection timeout” alerts because the message broker was unavailable.
The broker was in a “Memory Quarantine State”
There was a node started and required synchronising over 1.5M messages
Message broker underwent maintenance (restarts/replacement) with too many messages queued up
2 microservices with audit queues had stopped consuming messages.
...
Action Items
Action Items | Owner | Comments |
|---|---|---|
What happened? What caused the memory alarm? When was this an issue? Can AWS tell us anything about what was happening? | ||
Simulate the failure / test pre-emptive alerting | ||
Rearchitect:
| Mini-hack PoCs & improvements to python scripts: e.g. time-based scaling of audit consumers (based on when messages are usually published) | |
Alert on a certain queue size |
...