Slow Oracle RAC reconfiguration due to UDP packet reassembly failures
(Doc ID 2999850.1)
Last updated on JULY 20, 2024
Applies to:
Oracle Database - Enterprise Edition - Version 19.3.0.0.0 to 19.22.0.0.0 [Release 19]Information in this document applies to any platform.
Symptoms
Issue description
- RAC reconfiguration takes more than a minute to complete
- Most of the time is spent in SYNC steps
- packet reassembly failures are observed on all or some of RAC nodes during reconfiguration
- Similar performance issue will occur during "Dynamic Lock Redirection" operations
Applies to
- Any platform with UDP interconnect
- All DB releases without the Oracle bug-fix (BUG-35663859)
- System is not using jumbo frames for cluster interconnect, i.e. MTU size on the system is small like 1500
Changes
Example -- Over 400 seconds during PDB join reconfiguration
- The reconfiguration, 1-node → 2-node RAC (instance2 is joining), took 417 seconds.
- 406 seconds out of it was spent in sync 8 on instance1, and 414 seconds out of it was spent in sync 7 on instance2. (Ref: DLM RCFG HA STATS in lmon trace file).
- There were 8.6M fusion locks sent from instance1 to instance2 (Ref: DLM RCFG HA STATS in lmon trace file).
Cause
To view full details, sign in with your My Oracle Support account. |
|
Don't have a My Oracle Support account? Click to get started! |
In this Document
Symptoms |
Changes |
Cause |
Solution |
References |