dave@arnold.UUCP (Dave Arnold) (10/20/87)
Our company is dragging an under used CT Megaframe out of the closest, and going to run it as a dedicated news system :-) Since I am the one who suggested this, I have been appointed the job of administrating it. Also, I have been asked by upper management to prepare a proposal detailing how much time each week I plan to spend on this (I am a software engineer with projects and deadlines). I am not sure how much time WILL be required, so I thought I could pose this question to the experienced. Like I said, the hardware will be a Convergent Technologies Megaframe running CTIX, with about 80meg of disk space (probably 20-40meg free), and a 1200 baud modem, a telebit trailblazer looks good for the future. There are about 100 employee's who will be potential News readers/posters, most of them programmers. I recognize the following duties: 1. Create shell accounts as needed. 2. Provide Netnews user documentation as needed. 3. Provide technical support (Hey Dave, how do I send mail to somebody on DEC's Easynet?). 4. Monitor nightly newsfeeds, make sure everything ran smoothly, enough disk space, etc. 5. Fix problems with network/terminal servers to News system. 6. Review UUNET bill. That's about all I can think of. I have no idea how much time all of this is going to take, but I'll take a stab at about 2 hours a day? I would appreciate hearing from you old-timer USENET adminstator's, the full-timer's and the part-timer's, and hear your experiences/job demands. Thank you in advance. Dave Arnold -- Name: Dave Arnold USmail: 26561 Fresno, Mission Viejo, Ca, 92691 USA DDD: +1 714 586 5894 UUCP: ...!uunet!ccicpg!arnold!dave
jane@tolerant.UUCP (Jane Medefesser) (10/20/87)
In article <273@arnold.UUCP>, dave@arnold.UUCP (Dave Arnold) writes: > I recognize the following duties: > > 1. Create shell accounts as needed. > 2. Provide Netnews user documentation as needed. > 3. Provide technical support (Hey Dave, how do I send mail to somebody > on DEC's Easynet?). > 4. Monitor nightly newsfeeds, make sure everything ran smoothly, enough > disk space, etc. > 5. Fix problems with network/terminal servers to News system. > 6. Review UUNET bill. > > That's about all I can think of. I have no idea how much time all of this > is going to take, but I'll take a stab at about 2 hours a day? I administer the news at Tolerant part-time. Your guess is essentially correct. 2 hours a day should take care of everything you listed above. Actually, that's a pretty generous stab. The routine administration generally takes me about a half hour every morning. (monitoring disk usage, ensuring all the cron jobs ran exit 0, etc.) The nice thing about administering the news part-time, is that if you get really swamped at the office, you can set yourself up with a terminal & modem at home and do some of the nit-picky stuff from home. I think our installation is about average. We have 2 full feeds, one of which runs ihave/sendme. We have 1 full downstream site. We have local exchanges with about 4 other sites. Assuming that everything is installed correctly (THAT may take you some time...), once it gets running, it isn't terribly time consuming at all. (just news, that is!! Full time administration is another story!!!). FYI- my *real* job here is Software QA. QA is not terribly demanding - it's very flexible. I usually am able to drop whatever I'm doing to stomp out fires, then pick up where I left off. And, I am able to do A LOT from home when the need arrises. The two jobs compliment each other rather nicely. (*real* jobs are what you get paid for!!)
david@ms.uky.edu (David Herron -- Resident E-mail Hack) (10/20/87)
In article <273@arnold.UUCP> dave@arnold.UUCP (Dave Arnold) writes: >Like I said, the hardware will be a Convergent Technologies Megaframe running >CTIX, with about 80meg of disk space (probably 20-40meg free), and a 1200 baud >modem, a telebit trailblazer looks good for the future. There are about 100 >employee's who will be potential News readers/posters, most of them >programmers. 20-40 free megs is rather slim ... Unless this machine is just going to be a buffer between your internal machines and the rest of the world. (i.e. people don't READ news on this system, they read it on some other system ... this CTIX box can then run with a 1 week expire and be happy). We generally have 20-40 megs tied up here with news articles ... >I recognize the following duties: >1. Create shell accounts as needed. >2. Provide Netnews user documentation as needed. >3. Provide technical support (Hey Dave, how do I send mail to somebody > on DEC's Easynet?). [This is easy nowadays ... the gateways there rewrite the headers to say user@host.dec.com ... then you just get an intelligent mailer that knows the path for a gateway to dec.com and it happens automatically]. >4. Monitor nightly newsfeeds, make sure everything ran smoothly, enough > disk space, etc. >5. Fix problems with network/terminal servers to News system. >6. Review UUNET bill. [Weeell... there isn't much on the UUNET bill to review ... it'd be nice if they put out a listing of the calls you'd made, or possibly just the large ones (or non-prime-time) ones ... but then that's not a proper subject for here ... :-) ] hmmm ... You're putting E-Mail duties in there too. I work SOLEY as news/mail support for 20 hrs/week. The job has grown to the point where I now have a helper who I'm training ... This in a University where it's real hard to hire people to begin with, and in a support group which is fairly understaffed anyway. So we have 30 hrs/week (10 of which is a trainee remember) of work supporting our news/mail system and we're NOWHERE NEAR to catching up with what needs to be done. But then I've got a real complicated system here. Links to BITNET, Internet, UUCP and news being exchanged over all 3 networks. Different exchange methods used on all the 3 networks. Mail going across all 3 networks has different little idiosyncracies. Each of the subsystems I manage is as-large-or-larger than the kernal, and I have to make non-trivial changes sometimes just to get the subsystems to work. Your 16 hrs/week guess is a pretty good guess. Especially if you're new at administrating news/etc. -- <---- David Herron, Local E-Mail Hack, david@ms.uky.edu, david@ms.uky.csnet <---- {rutgers,uunet,cbosgd}!ukma!david, david@UKMA.BITNET <---- I thought that time was this neat invention that kept everything <---- from happening at once. Why doesn't this work in practice?
scott@helium.UUCP (Scott Hazen Mueller) (10/21/87)
In article <273@arnold.UUCP> dave@arnold.UUCP (Dave Arnold) writes: >Like I said, the hardware will be a Convergent Technologies Megaframe running >CTIX, with about 80meg of disk space (probably 20-40meg free), and a 1200 baud Well, I run in even tighter space, so I can throw in a couple of relevant comments. Your biggest day-to-day consideration is going to be disk space. If you feed anyone else, you need to be *very* careful about your space - it may be fine for your own use, but once/day compressed batches can really mess things up. >I recognize the following duties: > >4. Monitor nightly newsfeeds, make sure everything ran smoothly, enough > disk space, etc. Probably the most important thing. Most days, I don't even care what the numbers are. If you can get the Watcher program that was recently posted, get it and use it. I use it to monitor both total free space and the space used by the news directory tree. I actually don't *specifically* monitor the news feeds, but I do read news first thing in the morning, and so I'm aware that way of the news status. >That's about all I can think of. I have no idea how much time all of this >is going to take, but I'll take a stab at about 2 hours a day? On the two little systems that I run, 2 hours/day would more than cover *everything* - running news, installing patches, maintaining the pathalias database, etc. If you don't already, you'll want to read comp.sources.unix and news.*, which will tend to smooth out your time usage. >Dave Arnold \scott -- Scott Hazen Mueller City of Turlock 901 S. Walnut Rd. Turlock CA 95380 (209) 668-5590 lll-crg!csustan!helium!scott uunet!zorch!helium!scott
rbl@nitrex.UUCP ( Dr. Robin Lake ) (10/21/87)
In article <7540@e.ms.uky.edu> david@ms.uky.edu (David Herron -- Resident E-mail Hack) writes: >In article <273@arnold.UUCP> dave@arnold.UUCP (Dave Arnold) writes: >>Like I said, the hardware will be a Convergent Technologies Megaframe running >>CTIX, with about 80meg of disk space (probably 20-40meg free), and a 1200 baud >>modem, a telebit trailblazer looks good for the future. There are about 100 >>employee's who will be potential News readers/posters, most of them >>programmers. We've got about 12 users on a MegaFrame with CTIX and about the same disk space. With a newsfeed that is often irregular, most of the administrator's time is solving the problems incurred when running out of disk space due to the volume of incoming news. Yes, we're getting more disks, but that is another story.... If your newsfeed goes down, if your users tend to save a lot of interesting source or binhex'ed articles, if you ever really run a lot of data on the machine --- you better have a node to uucp archived or cpio'ed files onto! -- Rob Lake {decvax,ihnp4!cbosgd}!mandrill!nitrex!rbl
wcs@ho95e.ATT.COM (Bill.Stewart) (10/23/87)
In article <273@arnold.UUCP> dave@arnold.UUCP (Dave Arnold) writes: >Like I said, the hardware will be a Convergent Technologies Megaframe >running CTIX, with about 80meg of disk space (probably 20-40meg free) >and a 1200 baud modem. I spend less than 2 hours per week on news, but we have administrators who keep uucp happy and worry about disk space. Most of the time is in periodic hack attacks, when a new version comes out or when I update the pathalias maps. UUCP: We run Honey Danber UUCP from AT&T, which is *much* more reliable than old-uucp. We don't run sendmail, so we lose the functionality but also the headaches. Most of our uucp traffic comes over an AT&T Datakit packet network, which also makes uucp much cleaner. Disk: Get an extra disk! The reduction in workload will more than pay for your time. Remember that 24 MB/week will fill up disk fast, and your users will also want space. Make sure you use the same disk for news and /usr/spool. We take the approach that news is nice but expendable; if the disk fills up and I don't clean it up the other administrators expire stuff brutally. If the disk gets corrupted, news articles generally won't get restored; only libraries. - Keep an archive of comp.sources.* and any other groups that are relevant to your work, and make sure the users know about it. This will reduce the effect of all 100 users saving nethack and compiling it themselves. It's worth backing it up on tape periodically, since you probably can't keep it on disk for too long. -- # Thanks; # Bill Stewart, AT&T Bell Labs 2G218, Holmdel NJ 1-201-949-0705 ihnp4!ho95c!wcs
rob@philabs.Philips.Com (Rob Robertson) (10/23/87)
In article <273@arnold.UUCP> dave@arnold.UUCP (Dave Arnold) writes: >Like I said, the hardware will be a Convergent Technologies Megaframe running >CTIX, with about 80meg of disk space (probably 20-40meg free), and a 1200 baud >modem, a telebit trailblazer looks good for the future. There are about 100 >employee's who will be potential News readers/posters, most of them >programmers. it don't take that much time. 5 hours a week tops. i manage news here at philabs. 3 separate machines, with a user community of about 200. if you've got everything automated and know what you're doing (AND HAVE GOOD HARDWARE AND MODEMS), it's cake. 20-40 meg of disk ain't bad for a leaf site if its all reserved for transient news (ie it's in the spool directory, not the stuff the users have saved). we've got about 120 meg on our spool partition. we feed 14 sites, plus zillions of mail connections (the mail directory is /usr/spool/mail on bsd), and expire news after 14 days. we rarely break 60% capacity. things that add to your time: an unreliable news feed. if the machine feeding you goes down for a week expect to swallow 3 weeks of news two weeks after it comes up. feeding unreliable sites. at philabs that's one of the reasons we don't connect to home computers. if a disk goes out on a pc, it can be weeks before it's fixed. if it's a production machine (with the possible exception of the the world's largest computer company) it's up within days. the point being if you automate most of it with script, have good hard and software, and have a good news feed, you shouldn't have to spend much time at it. users are a different question..... good luck, rob -- william robertson rob@philabs.philips.com "better living through shell scripts"
zeeff@b-tech.UUCP (Jon Zeeff) (10/26/87)
I been running news on a system with limited disk space for quite awhile. Some things I have done to automate news maintenance: 1) An expire script is used that keeps expiring news until there is a certain amount of space free. A fixed period like 14 days is not used. 2) Outgoing news is limited to n MB/day. This prevents some of my neighbors from being swamped when things get backed up. 3) News is spooled "on the fly". A daemon process keeps checking the amount of news spooled and only spools more when it goes below a certain level. 4) There is a daemon that cuts off incoming uucp transfers when disk space falls below a certain level. News/uucp doesn't seem to handle no space very well. The combination of all these things makes it pretty hassle free. -- Jon Zeeff Branch Technology, uunet!umix!b-tech!zeeff zeeff%b-tech.uucp@umix.cc.umich.edu
rwhite@nusdhub.UUCP (Robert C. White Jr.) (10/27/87)
In article <3805@b-tech.UUCP>, zeeff@b-tech.UUCP (Jon Zeeff) writes: > I been running news on a system with limited disk space for quite awhile. > Some things I have done to automate news maintenance: > 2) Outgoing news is limited to n MB/day. This prevents some > of my neighbors from being swamped when things get backed up. > 3) News is spooled "on the fly". A daemon process keeps checking the amount of > news spooled and only spools more when it goes below a certain level. I have a unique solution for you... Make a file [like /usr/lib/news/makesp] with the following contents: /usr/lib/news/sendbatch -c ${UU_MACHINE:?'Bad Machine Name'} NOTE: UU_MACHINE may be called something else durring uuxqt on your machine. The person who wants their mail "uux"es the makesp on your system, waits a few minutes, and then calls back with some measealy request [if necessary] the program they run to pickup mail is something like: if uux not outstanding uux -s/usr/lib/news/LCK..sysname "sysname!makesp" if uux result positive sleep [some few minutes] /usr/lib/uucp/uucico -r sysname With this type of "mail pickup" script no spooling ever takes place for a system that is not responding. The remote site has to "request" their feed every time. The only standing disk usage is the "batch" file. Sleeping sites don't eat any disk space. I didn't include particualrs because this will vary site to site and remember that this is two connects, not one, but the savings are good for expensive hops. Rob.