Re: [squid-users] relation between page faults/sec and vmstat o/p

On Tue, 4 Sep 2001, khiz code wrote:

> hi all
> i was just wondering if ther eis any relation between the page
> faults/sec shown by cachengr and the vmstat o/p
> i hv benn getting something like
> client_http.requests = 18.046042/sec
> client_http.hits = 4.815610/sec
>
> unlink.requests = 3.709308/sec
> page_faults = 5.565629/sec
> select_loops = 188.571505/sec
> select_fds = 194.827005/sec
> average_select_fd_period = 0.005041/fd
> median_select_fds = 0.000000
> swap.outs = 4.782442/sec
> swap.ins = 4.979072/sec
>
> however vmstat shows me
> procs memory swap io system cpu
> r b w swpd free buff cache si so bi bo in cs us sy id
> 1 0 0 0 2212 200120 47428 0 0 1 4 147 67 1 2 98
> 0 0 0 0 2832 199656 47276 0 0 2 14 879 376 3 12 85
> 0 0 0 0 2372 199984 47404 0 0 4 25 760 334 2 9 89
> 0 0 0 0 2144 200092 47524 0 0 0 24 729 370 2 9 89
> 0 0 0 0 2972 199700 47140 0 0 1 13 616 306 2 7 91
> 0 0 0 0 2676 199892 47244 0 0 1 13 621 303 3 7 91
> 0 0 0 0 2224 200192 47396 0 0 2 13 645 316 3 14 83
> 0 0 0 0 2972 199912 46960 0 0 2 24 621 287 2 13 85
> 0 0 0 0 2600 200104 47140 0 0 1 17 642 303 1 7 91
> so do i need to worry abt this
> is there any relationbetween these stats.. vmstat shows me that no
> swapping is taking place
> admittedly i hv just 1 GB of cache_dir but 128 MB cache_mem
> but even with a 20 GB cache_dir one fine day that space has to get
> filled up and unlink comes into action.. isnt it

Its hard to say if they are related. It's going to depend on your
particular operating system.

the page faults reported by squid is just taken from the
getrusage() system call. So, you can read the getrusage()
man page, but thats probably not going to tell you very much
either.
Received on Thu Sep 06 2001 - 18:19:06 MDT