Date: Wed, 25 Jun 2003 16:29:36 +0400 (MSD) From: Varshavchick Alexander <alex@metrocom.ru> To: Paul Robinson <paul@iconoplex.co.uk> Cc: Lev Walkin <vlm@netli.lan> Subject: Re: How to delete unix socket entries Message-ID: <Pine.GSO.4.33.0306251614180.27984-100000@apache.metrocom.ru> In-Reply-To: <20030625115204.GI34365@iconoplex.co.uk>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 25 Jun 2003, Paul Robinson wrote: > Couldn't disagree with you more in general terms. Somebody else here (or on > -chat) once said something like "The fact that a server can stay up for 300 > days is a testimony to how good FreeBSD is. Unfortunately, it's also > testimony to how bad FreeBSD Systems administrators are". > > If you have a production system designed so that can't take one box out for > 120 seconds a week, your production system is wrong. You either need close > to something like 99.999% uptime, in which case you should be load-balancing > and/or clustering anyway, or you don't understand the problem. > > Patching your src tree, re-building and YES! Shock, horror! RE-BOOTING on a > regular basis is something a competent sysadmin is not scared to do. In > fact, not doing it just accepts that you don't care about the health and > security of your systems. You are prepared to let security holes seep in, > stale libraries clog up disk space, your machine gracefully degrades into a > horrible, horrible state. > > When I walk on site, if the sysadmin has a machine with an uptime over 150 > days and is proud of it, I know I'm going to have an uphill struggle to beat > a clue into their thick, stupid, ignorant heads. It's good how you expressed it here about the heads :) Paul, I agree with you about planned system upgrades and other activities, connected with the server being rebooted, and surely I _CAN_ reboot the server if it is _REALLY_ needed. The goal is to reduce the server downtimes as much as possible, and when it's really needed, because of a kernel upgrades or likewise and not for a fixing the consequenses of a broken applications. > > This is not about competing with MS. It's about running a secure, stable > environment. In your particular case, your environment isn't stable because > you have a server application that is core-dumping regularly. In that > eventuality (and seeing as you haven't given us a piece of code that > replicates the error) you have two choices: > > 1. Fix your broken server application Trying to do it. > 2. Reboot your machine on a regular > If this is the only way to make the server live when some applications are not behaving correctly, this is NOT a good thing. > This is the only way you will ever have your server in a continual known > state and operate productionally in the manner you want. If you cvsup > -STABLE and rebuild the world before a reboot, you also get the benefit of > performance security patches. This is a Good Thing. > Also you're getting new bugs and sideeffects going with the new kernel and world version, and need to rebuild and reinstall all or almost all of the software running on the server. This is of cause if you're keeping the server for performing some real work... ---- Alexander Varshavchick, Metrocom Joint Stock Company Phone: (812)118-3322, 118-3115(fax)
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.GSO.4.33.0306251614180.27984-100000>