我正在使用 PHP 的 google api 从我的服务器查询 google 的 API - 它使用 guzzle 发送常规 http 请求。大量请求 (~10%) 在 63 秒后返回，而其余请求在约 1 秒后返回。
我从另一台服务器尝试过，但该问题不存在。
我检查了 netstat，发现有大约 500 个时间等待连接：
netstat -ant | awk'{print $6}'| 排序| uniq -c | 排序-n
1 established
1 Foreign
4 CLOSE_WAIT
6 FIN_WAIT1
6 SYN_RECV
11 FIN_WAIT2
11 LAST_ACK
31 LISTEN
66 SYN_SENT
90 ESTABLISHED
456 TIME_WAIT
cat /proc/sys/fs/文件编号：
379008 0 500001
cat /proc/sys/net/ipv4/ip_local_port_range
24576 65534
我如何检查哪个设置导致响应如此延迟。
谢谢
网络状态监测
TcpExt:
26599 resets received for embryonic SYN_RECV sockets
147494 packets pruned from receive queue because of socket buffer overrun
158 ICMP packets dropped because they were out-of-window
1 ICMP packets dropped because socket was locked
1586409 TCP sockets finished time wait in fast timer
25630 packets rejects in established connections because of timestamp
2528562 delayed acks sent
2824 delayed acks further delayed because of locked socket
Quick ack mode was activated 160978 times
35706 times the listen queue of a socket overflowed
35706 SYNs to LISTEN sockets dropped
8532661 packets directly queued to recvmsg prequeue.
19911549 bytes directly in process context from backlog
3447934652 bytes directly received in process context from prequeue
60217414 packet headers predicted
4655144 packets header predicted and directly queued to user
80744319 acknowledgments not containing data payload received
42859898 predicted acknowledgments
26 times recovered from packet loss due to fast retransmit
184100 times recovered from packet loss by selective acknowledgements
292 bad SACK blocks received
Detected reordering 33568 times using FACK
Detected reordering 114073 times using SACK
Detected reordering 126806 times using time stamp
113848 congestion windows fully recovered without slow start
1980316 congestion windows partially recovered using Hoe heuristic
23835 congestion windows recovered without slow start by DSACK
25078 congestion windows recovered without slow start after partial ack
82939 TCP data loss events
TCPLostRetransmit: 4572
10 timeouts after reno fast retransmit
7421 timeouts after SACK recovery
4352 timeouts in loss state
838050 fast retransmits
108184 forward retransmits
87852 retransmits in slow start
182297 other TCP timeouts
3 classic Reno fast retransmits failed
4290 SACK retransmits failed
116 times receiver scheduled too late for direct processing
134674 packets collapsed in receive queue due to low socket buffer
164349 DSACKs sent for old packets
1551 DSACKs sent for out of order packets
775080 DSACKs received
1007 DSACKs for out of order packets received
153613 connections reset due to unexpected data
23137 connections reset due to early user close
58391 connections aborted due to timeout
TCPSACKDiscard: 33
TCPDSACKIgnoredOld: 10477
TCPDSACKIgnoredNoUndo: 457106
TCPSpuriousRTOs: 3193
TCPSackShifted: 670876
TCPSackMerged: 1268943
TCPSackShiftFallback: 661260
TCPBacklogDrop: 14555
TCPOFOQueue: 1173042
TCPOFOMerge: 919
TCPChallengeACK: 4644
TCPSYNChallenge: 432
TCPFromZeroWindowAdv: 17586
TCPToZeroWindowAdv: 17586
TCPWantZeroWindowAdv: 119758
TCPACKSkippedSynRecv: 4284
TCPACKSkippedPAWS: 11760
TCPACKSkippedSeq: 4434
TCPACKSkippedFinWait2: 107
TCPACKSkippedTimeWait: 249
TCPACKSkippedChallenge: 223
答案1
您的网络或应用程序出了问题，最好调整您的 tcp 设置或设置适当的 tcp 超时。以下是一些示例。
echo 30 > /proc/sys/net/ipv4/tcp_fin_timeout
echo 30 > /proc/sys/net/ipv4/tcp_keepalive_intvl
echo 5 > /proc/sys/net/ipv4/tcp_keepalive_probes