Date: Sat, 31 Aug 2002 14:57:47 -0400 (EDT) From: tonerboy <geoffrey@reptiles.org> To: "Jack L. Stone" <jackstone@sage-one.net>, <derek@durham.net> Cc: stable@freebsd.org Subject: Re: Updating world with least downtime Message-ID: <20020831141126.D90887-100000@iguana.reptiles.org> In-Reply-To: <3.0.5.32.20020831104657.014b5a00@mail.sage-one.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 31 Aug 2002, Jack L. Stone wrote: > At 11:05 AM 8.31.2002 -0400, tonerboy wrote: > >On Fri, 30 Aug 2002, Erik Trulsson wrote: > > > >> On Fri, Aug 30, 2002 at 03:40:31PM -0500, Jack L. Stone wrote: > >> > At 01:26 PM 8.30.2002 -0700, Brooks Davis wrote: > >> > >On Fri, Aug 30, 2002 at 03:07:14PM -0500, Jack L. Stone wrote: > >> > >> Yes, that's what I do after taking an image of the drive and also > using the > >> > >> "build box" as the first test to see if everything works okay. > Downtime is > >> > >> only a matter of seconds... riskier, yes, but acceptable with the > image to > >> > >> fall back on. Dropping to the single-user mode causes a downtime of > about > >> > >> 30 minutes otherwise on a 1.4G CPU machine... longer on slower > machines. > >> > >> Everthing is now done via scripts. > >> > > > >> > >If it's taking that long, it's got some serious problems even with the > >> > >rediculous post times many boards have today. My laptop (with 3400RPM > >> > >disk) takes <15. > >> > > > >> > >-- Brooks > >> > > > >> > > >> > I don't believe your laptop does installworld, mergemaster, etc all in 15 > >> > minutes.... we must must be talking different things.... > >> > >> Why not? My computer does a installworld+mergemaster in approx. 30 > >> minutes and it is an old 166 MHz Pentium with only 32 MB RAM (and with > >> a IDE controller that hasn't even heard of UDMA.) > >> > >> I would expect it to be significantly faster on a modern machine and 15 > >> min sounds quite plausible for a modern laptop with a somewhat slow disk. > >> > > My pII-233 took about 20 doing 4.6psomething (from april) to 4.6 > >Tuesday. Using mergemaster within the same major revision cuts down the > >grope and edit dramatically. Probably would have taken less if I hadn't > >walked away during the installworld. > > > > Which begs the question, what does this box do that it is uptime > >critical? I thought I heard firewall but may be mistaken. The beauty of > >FreeBSD is that it runs on commodity (read cheap) hardware. The highest > >zoot boxes of today can handle a wackload of connections before coming > >close to the wall. Why not take a look at load and consider replacing it > >with a couple of less than cutting edge boxes and rotate them in and out > >of service so you have a backup box you can rebuild at your leisure? > > > > And ummm .. Erik. I can build a kernel in under 20 minutes. The > >joy of 40 minute builds wore off quick (remember 2 hour builds with > >486s?). Buildworld in ~3 hrs. > > > > Cheers! > > > > When I replied to the orignal poster, whose main concern was about > minimizing downtime, my main focus was on explaining my own method used > which causes a downtime of only one reboot -- a few seconds on a production > box with 1GHz CPU and 7200rpm HDs. > Speaking of which, hope that is his email I've stuck beside yours. Nevertheless, this was posted to stable before I (foolishly evidently), tried to move this to private email (especially without looking back at who originally posted the question). My apologies in any case. And yes, you are quite correct. Whatever they are feeding the gerbils who churn the wheels makes for a might bit snappier boot. > I didn't realize this would become a "race" upsmanship discussion. In > looking back at the records (rather than off the top of head), my actual > downtime a few months ago when dropping to single-user mode for > installworld-mergemaster was about 20 mins and this was running mergemaster > -v and going through each change on a full-service server machine. > I've been using -svi and <<shudder>> dread the security level you must use to do that. Especially in light of the s'kiddie traffic I see trying to get into my network (which invariably traces back to new, unsecured installs). These are not exclusively IIS boxes (which are expected) but apache (granted usually linux but FreeBSD as well) & cobalt. I don't know how they are managing to send ssh from the same, but it certainly affects my sleeping. > Back to the original poster who hasn't replied further, there is a choice > of being down a few seconds or a few minutes depending on the machine's > speeds and approach. > And rather than having the highest zoot box, one CAN have redundancy for the same cost (2 p3 boxes ~= p4 cost, nay?). A little preparation and downtime approaches nil. That yardstick isn't too far off in any succeeding generation sets. > Now about the fastest beer guzzler....?? > Sorry. Even if I pour it in slowly over fifteen minutes, the box falls to pieces. Now should you be in this neck of the woods around UU time, some of my peers may offer to gladly accept your offer. In the mean time, I'll happily loft my tea tolling thumb in your honour (just don't tell it that's draft in the other hand please). Cheers! To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020831141126.D90887-100000>