Re: RAID 1

From: P.T. Breuer (ptb_at_oboe.it.uc3m.es)
Date: 12/30/03 

Date: Tue, 30 Dec 2003 08:30:26 GMT

Les Mikesell <lesmikesell@comcast.net> wrote:

> "P.T. Breuer" <ptb@oboe.it.uc3m.es> wrote in message
> news:o1fpsb.s23.ln@news.it.uc3m.es...
> > Les Mikesell <lesmikesell@comcast.net> wrote:
> > > Maybe it is the way leafnode is normally configured, but it isn't
> > > the way usenet software is supposed to work since the conversation
> >
> > It makes no difference [to whether I see earlier posts or not]
> > how the server is configured.

> Of course it does. It is entirely your choice how long the entire
> conversation remains at your fingertips. The man page shows
> 5 days as an example for how long to keep articles; the fedora

Stop focusing on the (leafnode) server - it is a straw man. I've said
so! If you like, assume that articles are kept for 5 days on the
server!

I really resent your ignoring the proper explanation! Do you want me to
give it again? You snipped it all away. Only to keep wailing about
servers, despite my saying many times that it's an irrelevancy. If you
want to see the (leafnode) config, it is

   ## Unread discussion threads will be deleted after this many days if
   ## you don't define special expire times. Mandatory.
   expire = 9

   ## Fetch only a few articles when we subscribe a new newsgroup. The
   ## default is to fetch all articles.
   #initialfetch = 100
   initialfetch = 200

   ## To avoid spam, you can select the maximum number of crosspostings
   ## that are allowed in incoming postings. Setting this below 5 is
   ## probably a bad idea. The default is unlimited crossposting.
   maxcrosspost = 5

   ## If you suffer from repeatedly receiving old postings (this happens
   ## sometimes when an upstream server goes into hiccup mode) you can
   ## refuse to receive them with the parameter "maxage" which tells the
   ## maximum allowed age of an article in days. The default maxage is 10
   ## days.
   maxage = 7

   ## maxlines will make fetch reject postings that are longer than a certain
   ## amount of lines.
   maxlines = 200

   ## minlines will make fetch reject postings that are shorter than a certain
   ## amount of lines.
   minlines = 2

   ## maxbytes will make fetch reject postings that are larger
   maxbytes = 50000

   ## timeout_short determines how many days fetch gets a newsgroup which
   ## has been accidentally opened. The default is two days.
   timeout_short = 1

   ## timeout_long determines how many days fetch will wait before not getting
   ## an unread newsgroup any more. The default is seven days.
   timeout_long = 14

   ## timeout_active determines how many days fetch will wait before re-reading
   ## the whole active file. The default is 90 days.
   timeout_active = 90

And as you can see, there is nothing radically strange about it.

> RPM installs with 20 configured, probably a more reasonable
> choice.

This not a reasonable choice at all - if there were about 600 posts a
day on aolm, for example (and I suspect there are), then 20 days would
be 12000 messages. Those 12000 messages would be in a single directory.
And you know the rest. As you can see from the config above and its
comments, maxage is 10 days by default, and RH as always provide a
b0rken config, one that will "hiccup" in this case. Mine is set to 7
days, which is more than the 5 days you posit.

> Your copy might be a broken version - I see this mentioned

My copy is a wonderfully working copy. It has been working perfectly
for more than 3 years.

 % ll /usr/local/sbin/leafnode
 -rwxr-x--- 1 news news 55352 Nov 13 2000 /usr/local/sbin/leafnode*

And it is on a machine that has been working perfectly since early 1995:

 % ll /boot/
 total 5086
 -rw-r--r-- 1 root root 130954 Jun 13 1996 System.map
 -rw-r--r-- 1 root root 204 May 25 1996 any_b.b
 -rw-r--r-- 1 root root 204 May 25 1996 any_d.b
 -rw-r--r-- 1 root root 512 Nov 6 1996 boot.0300
 -rw-r--r-- 1 root root 512 Nov 20 1996 boot.0800
 -rw-r--r-- 1 root root 4416 May 25 1996 boot.b
 -rw-r--r-- 1 root root 88 May 25 1996 chain.b
 -rw-r--r-- 1 root root 4279 Jun 13 1996 config
 -rw------- 1 root tec 65536 May 3 2002 map
 -rw-r--r-- 1 root root 192 May 25 1996 os2_d.b
 -rw-r--r-- 1 root root 413624 Nov 20 1996 vmlinuz.nopci
 -rw-r--r-- 1 root prof 465448 Apr 17 1997 vmlinuz.old
 -rw-r--r-- 1 root root 104166 Aug 23 1995 zSystem.map
 ...

> on the mailing list: http://wXw.lXaXnXdX.oXg/XrXhXvX/2X02/mXgX0444.html

Allow me to point that link at devnull. I did. Now people will have to
check the archives to find the link to the (leafnode) archives :-).

> > > context is useful. You are turning usenet into the party conversation
> > > of crowded room instead of the archived technical discussion it
> > > can be.
> >
> > No I am not! I am asking you to supply the relevant contextual
> > information to make head or tail of your post when seen on its *own*.

> Huh? WHAT? Can you repeat that?? I can't hear you... Note, that
> like you, it is my own choice not to see what you said, but it doesn't
> make for great conversation, does it?

I hear perfectly. It is your choice to mumble, if you mumble. I am
asking you not to mumble. No - I am telling you not to mumble, and
making you repeat yourself clerly if you do.

> > No it isn't. Do you have any problem understanding me? Or I you? No.
> > When I say that information is missing, it is missing. Not because I
> > say it is, but because it is. I am extremely literal minded, and I do
> > not invent tall tales.

> Yet, while you are cluttering the archives with unnecessary demands
> for repetitions or complaints about *where* people put the unecessary
> repeated context that also clutters the archives, others often answer
> the question as asked.

This is false. Or at least mendacious.

> > I have used the same software throughout all of usenet history. Well...
> > I started with "rn", and moved to "trn", and then went to "tin", and
> > stayed there. Or did I flirt with "nn" along the way? Hmmm..

> If you even glanced at the trn manual and used it for a few minutes
> on a server that kept a reasonable amount of history you would know
> that it can show and traverse a complete conversation any direction
> you like. I've barely touched tin but I think it does the same.

I'm sure it can do "the same", because I use it. But not while I am
reading a post. And not while I am replying to a post. It can do many
things, but exactly one of them at a time. Just as I can. Just as it
has always been.

> > This is an inevitable result of the way all client softare works.

> As it should. And it also provides an easy way to see the
> context of a post.

There you have taken my quote out of context. Let me take the
opportunity to restore it to sanity.

  As you can also see, immediately before reading post 9, which is your
  post, I have read not the previous post on this thread (post 3), but
  post 8, which is the newest post on the previous thread. And I have
  read 6 posts between the two even on this minimodel, at a time
  interval between them of 1 hour.
  
>This is an inevitable result of the way all client softare works.<

  So the effect is that I will have seen approximately 1000 posts between
  any two posts of yours on the same thread distanced by one hour, none
  of them relevant to your question (assuming something like 100 posts an
  hour per group on average, and 10 newsgroups read).

  I have a fairly good memory, but I see no reason why I should be
  expected to trust it. And I don't try. If your post does not contain
  the details necessary to understand it, that's your problem. All I do
  is tell you.

Don't try that again, or I'll lodge a complaint with your perversion
of posts comptroller.

> > So the effect is that I will have seen approximately 1000 posts between
> > any two posts of yours on the same thread distanced by one hour, none
> > of them relevant to your question (assuming something like 100 posts an
> > hour per group on average, and 10 newsgroups read).

See!

> > I have a fairly good memory, but I see no reason why I should be
> > expected to trust it. And I don't try.

> This is precisely why the reader software has a carefully designed
> mechanism to retrieve that information for you as you need it.

Uh - false. Or pass. It is not "precisely why". Rather, if it didn't
have the appropriate mechanisms, and people wanted them, it would not be
used.

But as it happens, I don't care, since I don't use any such mechanism.
I simply reply to posts as I see them, and I have already explained to
you that before seeing this post, I will have seen not the previous post
on this thread (yawn), but the latest post on the previous thread. I
saw your previous post on this thread about 16 hours ago, as I recall,
and since then I have read several thousand other messages, on other
threads, and I have just previously read the latest new message on the
thread immediately prior to yours. Let me show you again:

 That is
 if posts are as follows, arrranged in threads ordered by time of first
 post, and ordered timewise within posts, and X marks those that I have
 seen before, then the order in which I read them is:

   |-._
   | |_X
   | |_1
   | |_2
   |-._
   | |_3
   | |_4
   |-._
   | |_X
   | |_5
   |-._
   | |_X
   | |_6

 As you can see, this is "normal order" for reading all remaining posts.

 And the result of reading all posts in normal order, say in a new sweep
 through the newsgroup every hour, is that I read in this order,
 overall:

   |-._
   | |_1
   | |_4
   | |_5
   |-._
   | |_6
   | |_7
   |-._
   | |_2
   | |_8
   |-._
   | |_3
   | |_9

 As you can also see, immediately before reading post 9, which is your
 post, I have read not the previous post on this thread (post 3), but
 post 8, which is the newest post on the previous thread.

Is there something difficult to understand? It's a breadth-wise sweep
across the new leaves in the tree, every hour or so.

> > If your post does not contain
> > the details necessary to understand it, that's your problem. All I do
> > is tell you.

> So, just like the people who refuse to look through easily available
> archives to see if their question has already been answered, you
> clutter the network and archives with uninteresting gunk that no

You are verging on the argument you should be making. Go ahead and make
it, instead of supplying the above amateurish junk!

> one needs to see. Note that TCP networking works very well
> with no such thing as a negative ACK or request for a repeat. If
> you simply ignore things you don't understand the sender will
> repeat the request on his own.

I must say I think that is a valid socio-economic tactic. Hmm ...

Unfortunately while I do skip about 70% of posts because they're either
top posted or obviously unreadable before I even come to consider the
content, there are still about 30% of posts which look readable. Of
those maybe about 30% will be sufficiently badly and vaguely phrased
once one starts reading them for me to decide that the question is
easily answerable if the poster would only supply the relevant
information to make the post semantically coherent. You may consider
that the writer says in nicely laid-out text

    Yes, things are back to the way they were when I started now.

This merits a LART. He's mumbled. By now it's worth my while to tell
him so.

> > > and that contain the solution. How many of your posts are just
> > > clutter between the question and the real answer?
> >
> > About 25%, I imagine. It's easier to answer a message complaint than a
> > factual question, and impossible to answer a question that is
> > incorrectly framed (one will answer something else!). When the question
> > is in the correct form, then it can be answered.

> http://groups.google.com/groups?safe=images&ie=UTF-8&oe=UTF-8&as_uauthors=ptb@oboe.it.uc3m.es&as_scoring=d&lr=&num=100&hl=en

> Looks only a few have any useful content, let's say 10%. So if all

Stop making up the figures. It's a silly tactic. I don't know what that
link shows, but it seems to be a page of threads to which I have
contributed recently. As usual on google I can't see a damn thing for
their heinous interface, but it says:

 Searched Groups for author:ptb@oboe.it.uc3m.es. Results 1 - 100 of about 17,700. Search took 0.33 seconds.

> 17,700(!) posts you've made follow that pattern, you've made everyone

Let me see: on post 1 I answer his question (for the throd time, as I
recall!).

    Did you have some problem doing the man pppd /log thang? For
    the third time of asking ...

(he wanted to know how to log pppd messages to a file - it shows up
when you do what I told him to do).

On post 2 I answer his question

   man fetchmail.

(he wanted to fetch mail, and asked how - he didn't know about
fetchmail, and described precisely what it does in his question).

On post 3, I answer his question

   A thread waiting to do something. Get some more data on it. It
   probably doesn't expect to be sent to sleep/interrupted by apm
   events,

(he experiences random lockups when running Mdk9.1/2, and has printed
his process table, and I have pointed him at the kernel thread running
his connexant modem).

This is boring. I'm not going to struggle with googles html interface
any longer.

> who tries to find answers in the archives or the groups in realtime wade
> through 15,930 unneeded messages.

False. Do you really want to be made into a plonk?

> Now back to Grub...
> >> In any case the
> > > trick is to make the boot block on the 2nd drive (when installing) refer
> > > to itself and rest of the boot code on the primary bios drive because
> > > that is where it will be if it is ever executed. There are also some
> > > really ugly things that happen if you take the removed drive or a
> > > mirrored pair and swap it into a different machine that already
> > > has its own md setup, but that's a different issue.
> >
> > Yes - that's what I am essentially worried about. But also in the case
> > where you don't move the disk, but the machine changes. Which is
> > what happens when one of its disks disappears.

> No, with SCSI, both the bios and the linux device name mechanism
> will install the first working drive the same whether the broken one

That's the problem - the same bios code will now indicate a different
disk. Since you installed the boot sector when 0x81 meant the second
disk (i.e. 0x81 is referenced on the second disks boot sector), when
the sda breaks, 0x81 now indicates what was sdc, not what is now sda
(and was sdb).

You'd have to install the boot sector everywhere with references for
0x80.

How you do that and still retain the right physical offsets on the right
disks is lilo magic, and probably grub metaphysics.

> is removed or not - at least if the controller will boot at all. With IDE
> you normally can't boot without moving the drive and even if the

Well, without rejumpering, you mean.

> bios allowed it, a broken IDE drive typically keeps the controller from
> accessing the other drive on the same cable.

Yes. Often by pretending to be alive, but not being so!

Peter

Relevant Pages

  • Re: RAID Config for New Exchange 2003 Box
    ... Thanks Dave I greatly appreciate it. ... > backplane and drive cage will allow that config. ... >> likely the best to use with an exchange 2003 Server. ... >> layout the disk design. ...
    (microsoft.public.exchange.design)
  • Re: hardware recommendation
    ... This is a very reasonable config. ... I would schedule a job to archive those cubes after each full process, ... If we need more power will add some disks to share the workload> regarding the real usage of the server. ... It's just like you join some>>tables without using the index, you end up with disk scan. ...
    (microsoft.public.sqlserver.datawarehouse)
  • Re: hardware recommendation
    ... This is a very reasonable config. ... I would schedule a job to archive those cubes after each full process, ... If we need more power will add some disks to share the workload> regarding the real usage of the server. ... It's just like you join some>>tables without using the index, you end up with disk scan. ...
    (microsoft.public.sqlserver.server)
  • CEICW Fails at RegisterMSBOExchangeBP
    ... Ethernet adapter Server Local Area Connection: ... Call to Reading hardware selection returned ok. ... calling CNetCommit::ValidateRouterConnectionProperties. ... Call to Reading preferred DNS server IP returned ok. ...
    (microsoft.public.windows.server.sbs)
  • ICW Problem Error 0x80072581 Deleting the DNS record external NIC
    ... I am having major problems with the internet and email connection wizard in sbs 2004, curiously I have internet access however the wizard fails to complete during the network section when I try and add OWA and RWW web services. ... Server connects to the internet and provides access for server no clients attached as yet. ... Call to Reading hardware selection returned ok. ... Firewall Rule: SBS DHCP Client ...
    (microsoft.public.windows.server.sbs)