Re: [PATCH] MINOR: init: always fail when setrlimit fails

2019-10-22 Thread William Dauchy
Hello Willy, On Tue, Oct 22, 2019 at 04:46:43PM +0200, Willy Tarreau wrote: > To be honest, I'm quite embarrassed by such a change. I think the main > reason for the initial warning instead of error was that most people > using haproxy in unprivileged environments never knew what to use as > a

Re: [PATCH] MINOR: init: always fail when setrlimit fails

2019-10-22 Thread Willy Tarreau
Hi William, On Mon, Oct 21, 2019 at 11:14:20AM +0200, William Dauchy wrote: > On a production environment with a given maxconn, the fd limit can be > successfully set at a given time. While raising maxconn, a new max fd > limit is calculated. If the setrlimit fails (e.g if sysctl fs.nr_open > is

[PATCH] MINOR: init: always fail when setrlimit fails

2019-10-21 Thread William Dauchy
On a production environment with a given maxconn, the fd limit can be successfully set at a given time. While raising maxconn, a new max fd limit is calculated. If the setrlimit fails (e.g if sysctl fs.nr_open is lower), we silently revert to the max obtained through getrlimit, which can be way