Closed liguanquan closed 9 years ago
@wandenberg need your advice urgently
Hi @liguanquan sorry for the delay. I'm working to reproduce the problem. For now you can try to send small messages or use a common POST to publish the message.
ok i'll try split and join auto in program. thank you for this module again!!!
Hi @liguanquan
keep tracking the branch deal_with_ngx_again. The specific problem of publish a big message using WebSocket was solved. But as the message is sent to the publisher too, when using websocket, it keep being disconnected. I have to solve the problem on delivering big messages now.
greate it works !!! we'll do some loadtest.
Hi @liguanquan
how were your loadtests ?
its work fine in my production environment
Hi @liguanquan
can you repeat your tests with the code on fix_ngx_again_on_read_and_write branch (commit df6b37ae056b3a7bc97708e2380a0e6b7fdc5510). This version is equals the last one just updated with the code on master and with a fix for ngx_again when it happens on the write function.
Fixed with 312f330a01f92f7090a17d588209f970b25e86a4
hi wandenberg thanks your great work save me from openfire cluster my issue :i use server side websocket when message length > 3k connection close and message send fail nginx error.log give me client closed prematurely connection (11: Resource temporarily unavailable
client side websocket is fail too but longpoll is ok i update nginx and nginx-push-stream-module
sysctl.conf : 3 net.core.rmem_max = 8738000 4 net.core.wmem_max = 6553600 5 6 # increase Linux autotuning TCP buffer limits 7 # min, default, and max number of bytes to use 8 # set max to at least 4MB, or higher if you use very high BDP paths 9 10 net.ipv4.tcp_rmem = 819200 873800 8738000 11 net.ipv4.tcp_wmem = 819200 875360 6553600 12 14 net.ipv4.tcp_mem = 853440 1137920 1706880 15 16 net.core.rmem_default = 5000000 17 net.core.wmem_default = 5000000
give me some advice thanks