(1). Io server is full, the service is not responding,
sar -q -f / var / log / sa / SA28
The figure above shows plist-sz doubled
plist-sz Description: The number of processes in the list of processes (processes) and number (threads) thread
This algorithm parameters: number + the number of threads in all processes
All number of processes: ps axu | wc -l
Threads: pstree -p | wc -l
Extended:
Linux maximum number of threads and limit the current number of threads inquiry
1 summarizes the system has limitations:
/ proc / SYS / Kernel / PID_MAX maximum number of threads supported by the system # check, usually large, equivalent to the theoretical value
/proc/sys/kernel/threads-max
max_user_process # system limits the number of processes or threads can run at up to certain users, use the command: ulimit -u
Note: modify the value max_user_process, you only need to modify /etc/security/limits.conf, but the parameters need to be modified /etc/security/limits.d/90-nproc.conf
Appendix:
Appendix 1: centos 6. * can modify /etc/security/limits.d/90-nproc.conf
my side by modifying /etc/security/limits.conf, at last add
* Soft nproc 65535
* Hard nproc 65535
View default thread stack size, in bytes (Bytes), use the command: ulimit -s
/ proc / sys / vm / max_map_count # hardware memory size
2, Java virtual machine itself restrictions:
-Xms #intial the Java heap size
the -Xmx #maximum the Java heap size
-Xss the #The the Thread Stack size for the each
3, a program of the current inquiry the number of threads or processes
# pstree -p `ps -e | grep java | awk '{print $ 1}'` | wc -l
The above command is used to replace, on command substitution, that is bracketed by `` command will take precedence, then its output as arguments to other commands
or
# pstree -p process ID | wc -l
# Top -H process ID | wc -l
Used above is a pipe on the pipe: a pipe symbol "|" O command to the left as to the right of command
4, the entire system has been used to query the current thread or process the number of
pstree -p | wc -l
Number (2) io-threaded multi-lead to high disk read and write, resulting in service is not responding in a wait state
sar -d -f / var / log / sa / SA28
How many number of sectors wr_sec / s write per second
Description:
The total number of transmission I / O per second of: #tps
Total # rd_sec / s per sector read
Total # wr_sec / s per second write sector
# Avgrq-sz times per average disk I / O operation of the data size (sector)
# Avgqu-sz disk requests average queue length
#await complete request from the disk to the operating system processing, the average elapsed time for each request, includes a request queue wait time, in milliseconds (1 second equals 1000 milliseconds), the queue is equal to + seek time + service time
#svctm I / O processing time of service, i.e., not including the time of the request queue
#% Util I / O request CPU percentage occupied by the higher values, indicating slower I / O