Hello Jean-Yves, Jerome LAURET wrote: > > Hello Jean-Yves, > > Jean-Yves Nief wrote: > [...] > >> >> Thanks for looking onto this code. I don't see the exact bug fix for >> the network results, could you point me to it. > > > Boy. A diff would show it although I am not sure > I cna provide this for Pavel in the meantime. The main > thing is that the calculation around $measure[] was just > plain wrong, not retaining and rotating the value. There > is a if ($DEBUG) around there ... Also, I remember that > there was a pipe-like logic which would block for us > and not give results on SL 3.0.2 ... Jerome has already pointed out the problem but I would like to add more information. if !defined(open(CMDFD, "netstat -c -i $monint |")); I removed logic about pipe (line above) and getting the result each $monint by replacing while loop, which is sleeping $monint. Your logic caused to us some problems to get at least a result on SL 3.0.2. It was still waiting after the execution, sometimes it ran up after a long time, but it was very untrustworthy. The second thing about getting right network result. You used the array @meassure during the counting $net_activity: $net_activity = ($measure[1] - $measure[0]) * $mtu * 8 / $monint; But this array wasn't initialize anywhere and therefore the result was still 0. > >> as for paging I/O, I gave it up as it is highly correlated with other >> metrics such as memory occupancy. >> on top of that, retrieving the paging I/O is a bit tricky and I did >> not find any easy solution in the past for Solaris, plus it is hard >> to interpret (ie what is the max paging I/O activity one wants to >> accept). This is one other reason why I gave up with it :-). > > > > The values themselves do not matter much. Correlated > or not, the final formula is rather arbitrary (user's choice) > and Andy made this flexiblke enough that you may or may not > use it at the end. > > >> in your modification, you are looking at the % of swap usage not the >> I/O rate. You are taking the information from /proc/meminfo which is >> working for Linux, but not for Solaris or Mac OS. > > > While I beleive we have a if -e /proc/meminfo > ready to be epxanded for other OS as otherwise, pgio=0. > So, an imoprovement to the initial script while not perfect. > I hope you agree on the positive. > >> cheers, >> JY > > Cheers Pavel