Well, if you read archives it's a known problem with sticky-address. If you remove sticky-address everything works as expected. Following patch should work (I wrote it for 4.1 hoststated, not sure why it was not accepted, maybe quality of the code?, and modified it to work with 4.3 relayd), use it at your own risk.
Tautvydas > I have relayd in production with textbook (or manpage) examples > of doing redirects to backend webservers. > relayd seems to notice host state changes properly and makes > the necessary pf table adjustments. The problem is that the pf > states take so long to expire, that any clients that have existing > state table entries to the now-downed backend server continue > to talk to that downed server (I.e. they hang). > pfctl -k obviously resolves the issue, but shouldn't relayd kill those > states when it notices the host won't answer? > Thanks in advance. [demime 1.01d removed an attachment of type text/x-diff which had a name of relayd.diff]
