For me on 3.6 it occurred quite often, on 3.6.1 it happens quite rarely.
I am always on wired connection, so it happens there (as well?).
I have not recognized any particular pattern, just happens when switching to another menu item as mentioned earlier.
Unfortunately the “dmesg” logs are run out of that time, I only find this:
first time when the error is logged by lighttpd (when the menu navigation is broken):
2017-03-18 07:30:20: (connections-glue.c.291) connection closed - read failed: Function not implemented 38
the following is also logged (ignore the -1 hour diff, that is irrelevant):
2017-03-18T07:30:16+01:00 warning dnsmasq-dhcp[16334]: DHCP packet received on eth1 which has no address
2017-03-18T06:30:20+01:00 warning : Unable to use //tmp//yin2yang.xsl (No such file or directory).
2017-03-18T06:30:20+01:00 warning : YANG format data models will not be available via get-schema.
2017-03-18T06:30:20+01:00 warning : Unable to get tty (Not a tty) to get the client’s hostname (session 15750).
2017-03-18T07:30:40+01:00 warning dnsmasq-dhcp[16334]: DHCP packet received on eth1 which has no address
I do not know, if they are related or not.
Now I could not reproduce the issue explicitly, but I do not apply the “setenv.set-environment” fix, to be able to collect additional info if it happens again.
I used wireless connection several months, never happened to me. Now I use wired connection for about 1 month and I’ve had only one lighttd issue so far. Problem occurred 2 hours after restart on 2GB model. I have no messages in dmesg about the same time whatsoever (last message is in time of restart, ie 2 hours ago lighttd crash). There is nothing suspicious in system.log (messages) too. The only message around that time is warning watchdog[]: Restarted nethist but it is something occurring every 10 minutes. I use material theme for LuCI and some additional collectd mods and I have 2 not bridged VLANs in my network
If DHCP is unstable, and interfaces drop addresses, it is quite possible that firewall connection tracking and lighttpd listen sockets could be affected. Thanks. This gives me an avenue to try and reproduce the issue. Are there any other services listening on the box besides lighttpd? Do they have similar issues to lighttpd when dnsmasq reinitializes network interfaces?
May be that dnsmasq about DHCP are related, though I do not really see the reason.
My log is/was always quite full of such warnings in general in all the time (around 20 times in every 5 minutes), eth1=wan.
All other services were working fine, I mean openvpn, dlna, transmission, samba, etc.
Nikola log monitor were successfully executed right before:
2017-03-18T07:30:09+01:00 info nikola: (v41) Sending records took 0.044554 seconds
I rather thought
2017-03-18T06:30:20+01:00 warning : Unable to use //tmp//yin2yang.xsl (No such file or directory).
has something to do with the issue, but I now see it is explained here, as well I got that reguraly.