maximum number of FD events

[Cathy Almond]

On 25/10/11 21:09, Fr34k wrote:
> 
> 
> Hello,
> 
> Environment:  Solaris10 SPARC and x86, BIND 9.7.3-P3 and 9.8.1
> 
> Anomaly:  In our logs, we have been noticing "maximum number of FD events" entries.  For example,
>   named[8592]: [ID 873579 daemon.info] sockmgr 288760: maximum number of FD events (64) received
> 
> Action:  Our web searches have not found a lot of information on this.  One resource suggested that a few of these are "normal".
> 
> We found someone else who seemed to suggest a "fix" by increasing the number of sockets.
> We figured we would give that a shot and see what would happen.  We tried 128, and then 256 -- but we still see these messages:
>   named[14050]: [ID 873579 daemon.info] sockmgr 288760: maximum number of FD events (128) received
>   named[15910]: [ID 873579 daemon.info] sockmgr 288760: maximum number of FD events (256) received
> 
> Does anyone have more information, suggestions, comments?
> 
> Thank you.

This can be 'normal' if you are just seeing the log message
intermittently.  It just indicates a peak in socket activity.  Nothing
should be dropped as a result - it's simply that when named's i/o
watcher polled for socket events, there were more than anticipated so
only the first 64 could be notfied.  The remaining events should get
picked up next time around.

If the message persists, then there is probably something else 'going
on' that you need to investigate.

https://deepthought.isc.org/article/AA-00508/