[oracle@ ~]$ netstat -s
Ip:
809822061 total packets received
55926 with invalid addresses
0 forwarded
0 incoming packets discarded
466370724 incoming packets delivered
444574924 requests sent out
45010 fragments dropped after timeout
413215646 reassemblies required
70404407 packets reassembled ok
450201 packet reassembles failed
55591164 fragments received ok
315244982 fragments created
Icmp:
34512 ICMP messages received
5 input ICMP message failed.
ICMP input histogram:
destination unreachable: 19780
echo requests: 14686
echo replies: 46
49642 ICMP messages sent
0 ICMP messages failed
ICMP output histogram:
destination unreachable: 2612
time exceeded: 32290
echo request: 54
echo replies: 14686
IcmpMsg:
InType0: 46
InType3: 19780
InType8: 14686
OutType0: 14686
OutType3: 2612
OutType8: 54
OutType11: 32290
Tcp:
32684 active connections openings
40302 passive connection openings
45 failed connection attempts
1220 connection resets received
260 connections established
82554713 segments received
84688238 segments send out
991 segments retransmited
5 bad segments received.
240 resets sent
Udp:
382559874 packets received
5278 packets to unknown port received.
0 packet receive errors
359836050 packets sent
TcpExt:
64 invalid SYN cookies received
36429 TCP sockets finished time wait in fast timer
18 time wait sockets recycled by time stamp
1367903 delayed acks sent
428 delayed acks further delayed because of locked socket
Quick ack mode was activated 874 times
75519212 packets directly queued to recvmsg prequeue.
7538845 packets directly received from backlog
2489004686 packets directly received from prequeue
929028 packets header predicted
64341161 packets header predicted and directly queued to user
1451291 acknowledgments not containing data received
69929348 predicted acknowledgments
98 times recovered from packet loss due to SACK data
TCPDSACKUndo: 122
140 congestion windows recovered after partial ack
125 TCP data loss events
81 timeouts after SACK recovery
132 fast retransmits
19 forward retransmits
124 retransmits in slow start
564 other TCP timeouts
1 sack retransmits failed
2 times receiver scheduled too late for direct processing
874 DSACKs sent for old packets
260 DSACKs received
2 connections reset due to unexpected data
28 connections aborted due to timeout
IpExt:
InMcastPkts: 3994
OutMcastPkts: 45
InBcastPkts: 1215314
[oracle@ ~]$ more /proc/sys/net/ipv4/ipfrag_low_thresh
393216
[oracle@ ~]$ more /proc/sys/net/ipv4/ipfrag_high_thresh
524288
[oracle@~]$ more /proc/sys/net/ipv4/ipfrag_time
30
基本可以确定系统性能瓶颈为网络问题和SQL问题。
问题是：网络这块重组成功比率是多少可以接受？
如果要优化，怎么优化？
