Date: Wed, 19 Jun 2002 10:59:26 +0200 From: Andy Sporner <sporner@nentec.de> To: Derek Barrett <derekbarrett@graffiti.net>, freebsd-cluster <freebsd-cluster@FreeBSD.ORG> Subject: Re: Application cluster Message-ID: <3D1047EE.4000505@nentec.de> References: <20020618172808.25913.qmail@graffiti.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Hi Derek, >hahahahaha well as a fellow American then I should >have replied, "Thanks partner! USA!" > Well I am relieved to see that this is still something that exists! Today I just left the main FreeBSD hackers list because I found it to be just a margin too cut-throat. These days with what little time I have the last thing I want is not to be taken seriously. I regard email as a primary communications method and for some to just pretend the mail never got there is hard to tolerate (especially when I send it directly to the individual--2-3 times). I have the feeling that people are too childish to face things directly. Which is one of the reasons I am working here--better environment. -- but enough flaming for the moment ;-) > >I don't think you should dismiss your scripts that >"only start and stop" as being laughable. To me, >that's 75% of the battle. I know I've spent hours >at times just getting my startup scripts to work >properly, missing a switch here or there, the trial >and error involved in that is alot sometimes. And >getting a RELIABLE method of monitoring the other >servers has still been a challenge for everyone. > Thanks for your complement... > >Truly, getting a failover >server to successfully take over means: > >1) Reduced late night phone calls >2) Not having to make as many late night phone calls :-D > I used to work at Hyatt's central computer division in Chicago and I had many times "pager duty" and, yes I can sympathize with you! > >And most of these types >of scripts depend on having a second network card >and a serial cable as well. The Linux HA >servers even have a controlling server for the entire >cluster called a Director. That your mechanism goes >across a network card is nice, the less overhead, the better. > I allow the configuration of network addresses for each node. A heartbeat message is sent out over any and all links that are present for the server. The whole thing with the serial cable seems rather archeaic. I mean if the networking layer has failed, the server is probably not that usefull anyways! > >I mean, a couple thousand dollar hardware failover solution >is nice, but so would a Ferrari as a company car. I recently worked >in a high uptime enviornment, and every single server there had >an identical backup, run by a hardware failover switch, and >let me tell you, I got really SPOILED. The amount of >stress relief that those failover switches provided made troubleshooting >and maintenance a breeze. > Funny thing, I worked for about 5 years with Sequent clusters and in their earlier versions (< 2.0) the stand-alone machine was more reliable that the same machine in a cluster. At that time they really never got more than 2 nodes working right. That was why I started with 3 nodes in the beginning as it changes the dynamics remarkably and these same dynamics work very well on 2 nodes too. > > >Let me see what I can come up with for a place for you to post your file. > Meanwhile I need to dust off the work I was doing (I had modularized it with DSO support and added some process statis collecting so that you can from one point monitor processes on any node of the cluster). I also improved the build environment--it was previously very rickety. Andy To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-cluster" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3D1047EE.4000505>