Symptom
ISIS memory leak was observed on continuous route churn and interface flap
Conditions
If the there is a local topology change (interface up/down, etc.) in the network then this leak will be observed.
The leak is in the order of 100 bytes each time there is a local topology change.
Workaround
There is no workaround that prevents the leak. The leaked memory is recovered by an isis process restart.
Further Problem Description
ISIS is configured with multiple or single process. Database can be just L2 or L1 or L1/L2.
Due to some changes in the network the isis database is update and routes get churned. Over a period of time there is a possibility of leak because of this.
This fix helps mitigate the issue.
The bug has existed since 6.7.2 and affects all platforms
A memory leak in isis will only affect the isis process on the RP. ISIS normally uses less than 400MB. If the leak is severe enough the isis process will be restarted when the process reaches 2GB. The restart should have minimal impact on the network as long as "nsf cisco" or "nsf ietf" is configured and it will free the leaked memory.
