Troubleshooting - a bunch of examples of real world issues and how they were identified and fixed
summary
Public OpenStack clouds show two concrete fixes: a Heat stack becomes undeletable when Nova rejects user-data larger than 64KB and Heat then tries to store Nova’s 130KB error into MySQL’s text status reason overflow, fixed by trimming ~20KB from user data and filing a Heat bug to truncate/expand the field. After a Victoria→Sienna upgrade intermittent API failures and Galera aborted connections trace to keepalived VRRP packets delayed; tcpdump, HAProxy and MySQL logs reveal non-unique VRID across L2, resolved by making VRIDs unique and stopping keepalived on backups while reporting the keepalived regression.
AI-written summary · about
chapters (16)
- 0:00Intro
- 0:20Setting the stage
- 1:18Finding issues
- 2:161: Initial report
- 2:40Heat template
- 2:51Client side error
- 3:16Server side logs
- 4:08Limits
- 4:52What's actually going on here?
- 6:12Cause - two separate issues
- 7:462: Initial report after upgrade from V- X
- 10:23Focus on db
- 13:52Workaround
- 14:48Focus on keepalived
- 16:13Historical reasons
- 18:43Resources for troubleshooting
related talks
21mI Don’t Think This Means What You Think It Means: Red Herrings in OpenStack
22mMySQL and OpenStack
16mAuto-scaling in Openstack without Telemetry
30mOpenStack Works, so now what?
27mLessons Learned from a Large Scale OpenStack Deployment with Tripleo
03mRecap
01h 50mKeynote
23mOpen Matters: HP & OpenStack