I am counting current open files by a process pid by :-
cmd1:-
ls /proc/$pid/fd/* | wc -l

and then i am calculating percentage by fetching Max open Files limits.
cmd2:-
cat /proc/pid/limits

Limit Soft Limit Hard Limit Units

Max resident set unlimited unlimited bytes
Max processes 30425 30425 processes
Max open files 4096 4096 files
Max locked memory 65536 65536 bytes


Only pasting part of output.

I am calculating percentage :-

cmd1/cmd2*100

i.e if ls /proc/$pid/fd/* | wc -l gives 1000.
then percentage will be :-
1000/4096*100= approx 25%

sometimes i am getting this percentage as 220 in logs.
scenario is not reproducible now,it happened in some server.

It means process pid is crossing the Max open file limits(4096 ) .

First: /proc/<pid>/fd is only showing file descriptor links to regular files not ALL open "files".
That is to say in UNIX/Linux nearly everything is treated as a "file" not just the regular files for example you might see shared libraries open in memory, sockets (network connections on ports) and devices. To really see every "open file" for a process a much better tool is lsof. If you run "lsof -p <pid>" it will show you everything in use including your file descriptors (typically as CHR entries).

Second: The limits you see are both hard and soft. For root processes the soft limit can be exceeded to hard limit. For other users the ulimit command can be used to set limit for that invocation up to the hard limit but will default to the soft limit otherwise.

Although I've seen many things hit the max open files limit I've never seen anything actually exceed it.

We were using lsof command earlier ,but lsof command was tawas taking longer time to finish when /mnt/ was down .
so we started using ls /proc/pid/fd.
So, i just want to know whether limit can be crossed when i use ls /proc/pid/fd ?