server.event-handler = "linux-sysepoll"
server.network-backend = "writev"
When using writev under high load (over 500 simultaneous connections) serving large static files, I get tons of these messages in error.log:
2008-02-25 21:33:08: (network_writev.c.303) write failed: Bad address 795
2008-02-25 21:33:08: (connections.c.603) connection closed: write failed on fd 795
I've seen this:
http://trac.lighttpd.net/trac/changeset/829
and this bug is supposed to be fixed, but it's not.
Users are complaining that when they use download managers they fail to download the last segment of file (ie download stops at 99%). We were able to reproduce that, when trying to download really small chunk (740 bytes) from the end of 500mb file, lighttpd resets the connection (download manager gives "error occurred!" messages without transferring anything) without any HTTP headers sent.
Please fix this as writev is basically the only backend capable of serving large files. sendfile() causes lightly to work vey slow at 200+ connections due to high iowait.
