Detection of packet processing slowdown
Sometimes Kea will find itself unable to keep up with incoming packet load for various reasons (overwhelming amounts of packets in an avalanche scenario, slowdown in SQL queries in the case of database usage, and etc). Kea currently has no way to detect or warn about this situation. Detailed analysis of the logs is necessary to understand what is happening. This issue is meant to provide some ideas for future development in this area where Kea could possibly detect this situation and provide warning log messages about it.
Possible ideas:
- Add timestamps to received packets (possible? perhaps) as they are put into the buffer. Check these timestamps against the current time as they are pulled out of the buffer. If there is some discrepancy that is larger than some threshold, emit some kind of log message about this.
- In
netstat -l
output, there is a representation of the current buffer size for a process. Example:
$ netstat -l
Active Internet connections (only servers)
Proto Recv-Q Send-Q Local Address Foreign Address State
...
udp 0 0 10.1.2.2:bootps 0.0.0.0:*
Administrators sometimes use this to find out if some process has a slowdown in processing packets. If the number is larger than normal, then there might be. Kea may not necessarily know what is normal, but there must be some maximum size of the Recv-Q. If so, perhaps this maximum size is being reached when these backups occur. If that could be detected, perhaps a warning message could be logged.