Hello ppl, I've mailed before about some problems we currently have with a loaded Linux 2.4.1 server. We got mysterious errors but most of them went away when we replaced our switch. However, now when we run at full speed on the network we see the following repeating endlessly in our syslog. This is @ approx 6000-7000 incoming client connections to our IRC server. Feb 12 18:23:09 mcquack kernel: Out of socket memory Feb 12 18:23:09 mcquack kernel: TCP: too many of orphaned sockets Feb 12 18:23:09 mcquack kernel: TCP: too many of orphaned sockets Feb 12 18:23:20 mcquack kernel: NET: 2 messages suppressed. Feb 12 18:23:20 mcquack kernel: TCP: too many of orphaned sockets Feb 12 18:23:20 mcquack kernel: TCP: too many of orphaned sockets Feb 12 18:23:30 mcquack kernel: NET: 4 messages suppressed. Feb 12 18:23:30 mcquack kernel: TCP: too many of orphaned sockets and so on... Then we did the following tuning attempt: echo "20480" > /proc/sys/net/ipv4/tcp_max_orphans echo "1" > tcp_orphan_retries echo 30 > /proc/sys/net/ipv4/tcp_fin_timeout echo 4 > /proc/sys/net/ipv4/tcp_synack_retries echo "10000 49152" > /proc/sys/net/ipv4/ip_local_port_range echo 30 > /proc/sys/net/ipv4/tcp_keepalive_time echo 2 > /proc/sys/net/ipv4/tcp_keepalive_probes And yes, the box *might* be more responsive on the net but the errors keep coming? What can we do? Tuning? Is this a kernel problem? Oh btw. we got what I think is a "bad one" too... Feb 12 16:42:35 mcquack kernel: KERNEL: assertion (tp->lost_out == 0) failed at tcp_input.c(1202):tcp_remove_reno_sacks Regards Magnus Walldal
BEGIN:VCARD VERSION:2.1 N:Walldal;Magnus;;IT-Director FN:Magnus Walldal ORG:Business Linc i Göteborg AB ADR;WORK;ENCODING=QUOTED-PRINTABLE:;;Kungsportsavenyn 18=0D=0AG=F6teborg 41136=0D=0ASweden LABEL;WORK;ENCODING=QUOTED-PRINTABLE:Kungsportsavenyn 18=0D=0AG=F6teborg 41136=0D=0ASweden EMAIL;PREF;INTERNET:magnus.walldal@b-linc.com REV:20001005T142120Z END:VCARD
