We are observing the following behaviour with build 1.0.0-27828 (and earlier versions):
Depending upon the number of network connections (the apache vm is getting roughly 100k pageviews per day):
\- network connections from outside start to get lost
\- the load of vmnet-natd goes up from 1-5% to 30-40%
\- first only a few then all pages will no longer be served
The problem vanishes completely after manually killing and restarting vmnet-natd on the host system. The client is not touched at all.
We are now hourly restarting the nat process. This solved the problem but is very unsatisfying.
Has anyone observed a similar behavior or found a solution for it?
Regards,
Juergen