A large number of failed to acquire scoreboard appear in the PHP-FPM log

PHP-FPM Log In a large number of failed to acquire scoreboard, resulting in the child process to the maximum number of requests, kill, will not pull up a new child process, finally, the master process is left.
The logs are as follows:

ERROR: [pool ] no free scoreboard slotWARNING: [pool www] child 31311 said into stderr: "WARNING: failed to acquire proc scoreboard"

After troubleshooting, at the beginning of this problem, a request appears 500, 500 of the situation is due to the query database result set is relatively large, and there is this situation, the system log also appeared

TCP: time wait bucket table overflow

Do you know the reason?

Reply content:

PHP-FPM Log In a large number of failed to acquire scoreboard, resulting in the child process to the maximum number of requests, kill, will not pull up a new child process, finally, the master process is left.
The logs are as follows:

ERROR: [pool ] no free scoreboard slotWARNING: [pool www] child 31311 said into stderr: "WARNING: failed to acquire proc scoreboard"

After troubleshooting, at the beginning of this problem, a request appears 500, 500 of the situation is due to the query database result set is relatively large, and there is this situation, the system log also appeared

TCP: time wait bucket table overflow

Do you know the reason?

WARNING: failed to acquire proc scoreboardis not due to the large data result set query data are put in memory operation, and then the server system memory is insufficient, causing this problem.
time wait bucket table overflowBecause of a problem with processing requests, a large number of TCP connections are made, and the number of connections exceeds the maximum value of the system configuration tcp_max_tw_buckets .
Try the workaround:

1. Increase system memory

2. Block processing of large data result sets, or read large database result sets by rows (cursor mode)

3. Modified tcp_max_tw_buckets related values

The above is purely personal speculation, for reference only, because I have not encountered this error

Check to see if any of the scripts consumed a lot of memory.

