|
Grex > Coop11 > #36: Increasing User Disk Space Maximum | |
|
| Author |
Message |
| 23 new of 111 responses total. |
rtg
|
|
response 89 of 111:
| 
Feb 8 07:57 UTC 1999 |
I don't recall hearing a critical point. Does Grex' Sun/4 have
differential drivers on it's SCSi interface?
|
steve
|
|
response 90 of 111:
|
Feb 8 15:45 UTC 1999 |
We have some adaptors which will let us use differencial SCSI
disks on a standard single ended controller, and the 'new' 4/670
CPU board can talk differencial directly.
|
janc
|
|
response 91 of 111:
|
Feb 9 02:53 UTC 1999 |
Actually, the new motherboards are just like the old ones, and the SCSI driver
on them is not differential, but we have gained two new SBus cards containing
differential SCSI controllers which can be plugged into any of our
motherboards.
|
steve
|
|
response 92 of 111:
|
Feb 9 03:30 UTC 1999 |
Huh? We have *two* differencial SBUS scsi controllers? Wonderful!
|
janc
|
|
response 93 of 111:
|
Feb 9 06:48 UTC 1999 |
I believe so. One on the anonymous motherboard, one on the SBus
expansion card.
|
mdw
|
|
response 94 of 111:
|
Feb 9 12:09 UTC 1999 |
That means we could also have multiple scsi busses, a possible win.
|
senna
|
|
response 95 of 111:
|
Feb 11 11:47 UTC 1999 |
Pardon my ignorance, but have we just tabled the mail machine idea? It
seems like spare disks would be a wonderful starting point. We have
some extra processors running around, too.
|
steve
|
|
response 96 of 111:
|
Feb 11 12:04 UTC 1999 |
It isn't the lack of disks thats preventing us so much as the
lack of staff time. Also, the disks we have are all SCSI and the
mail machine hardware is IDE at this point. The mail machine
doesn't need that much disk, really. If it has 100M of disk
buffer space, we'll be fine for at least a solid day of it catching
mail in the circumstance of Grex being down. I forget the size
of the disk on that machine right now but I think we have a couple
hundred meg free on it at the moment.
|
tpryan
|
|
response 97 of 111:
|
Feb 12 03:31 UTC 1999 |
I have a couple of 160M IDE drives available from those PS/2
machines I've been rescueing and repairing. May or may not be what is
what you are looking for. Let me know.
|
mdw
|
|
response 98 of 111:
|
Feb 12 03:54 UTC 1999 |
Actually, one of the main reasons I haven't put more time into the mail
server, is I don't think we have a really pressing need for it yet. So,
rather than put time into sendmail on another machine, I've put time
into putting better anti-spam fangs into sendmail on grex, and into
kerberos for grex. This does have implications for the mail machine.
When we do deploy the mail machine, it really should have similar
anti-spam fangs as grex does -- otherwise, what will happen is that all
the spam will be accepted by the mail machine, then bounced by grex.
Kerberos has even more interesting impliations -- it could be used to
facilitate having the mail machine actually store, as well as collect
mail - which would allow us to offload much of the mail processing
completely off grex. It is also possible that we might take the machine
that had been planned as the mail machine and use it for kerberos, at
least temporarily.
|
steve
|
|
response 99 of 111:
|
Feb 12 04:10 UTC 1999 |
PS/2's that used IDE disks? Thats interesting.
|
devnull
|
|
response 100 of 111:
|
Feb 12 07:03 UTC 1999 |
Re #98: If the incoming mail is queued on the mail machine, does this mean
you're planning to do MIT-style POP, or something else?
|
mdw
|
|
response 101 of 111:
|
Feb 12 09:26 UTC 1999 |
Not sure what you mean by "MIT-style" POP. Do you mean POP2/POP3? or
KPOP?
We would probably choose to do a "special" version of KPOP. It would
only work here with the mail clients we support. Grex already has quite
enough "free email" traffic as it is; the last thing we need is to be
overwhelmed by a flood of people using the various popular pop clients
at home.
|
devnull
|
|
response 102 of 111:
|
Feb 13 04:11 UTC 1999 |
MIT uses some sort of Kerberized POP; I'm unsure of the details, and
don't use it myself (I have my MIT mail forwarded to GNU).
|
mdw
|
|
response 103 of 111:
|
Feb 13 10:30 UTC 1999 |
I use "kpop" at work, with k4 (des). This may be the same thing as MIT;
the oldest name I could find on what we use is Dave Bachman (formerly
<dave@citi.umich.edu>, then <wk01200@worldlink.com>.)
|
devnull
|
|
response 104 of 111:
|
Feb 13 22:49 UTC 1999 |
Then there are also krb4 vs krb5 issues, which I don't completely
understand. Apparently, zephyr still requires krb4; I think telnet
connections at MIT use krb5 these days.
|
senna
|
|
response 105 of 111:
|
Feb 16 12:12 UTC 1999 |
That makes sense to me. I recall the mail machine becoming a large issue back
before we had the 670 up and running and people would idle out waiting for
pine to load. The only real benifit I can see to a mail machine's capacity
is the potential for increasing ttys, which are always a hot commodity.
|
steve
|
|
response 106 of 111:
|
Feb 16 12:48 UTC 1999 |
Ah, 'mail machine' in which sense? There are two that have been
talked about. I think the staff has more thought along the lines of
a machine which would be the primary mail-catcher, such that Grex
itself did not spawn off sendmails constantly to get the mail. In
this system there would be one near enternal smtp connection between
the mail machine and Grex, which would relieve a burden from Grex.
The other idea would be to have a seperate machine for mail that
people use. I like this idea much less, because it has implications
for maintenance, and basically makes more work to do.
|
srw
|
|
response 107 of 111:
|
Feb 18 21:16 UTC 1999 |
The mail machine, as currently proposed, is only to catch smtp mail from
the internet and feed it to grex one message ata time. This will make a
positive impact on Grex. This machine is not a post office machine. I do
agree that the sendmail fangs Grex currently has, while not sufficient
yet, are a minimum starting point for what the mail machine would need.
I think it would be useful if we set up such a machine but I also think
it would be useful to move to a kerberos-based distributed services
environment. This could potentially all for distributing other
services besides mail. This is a much longer-range project, though.
It's a tough call whether to spend time on a short term or long term
project, but if Marcus doesn't want to spend time on the short-term one
because he wants to spend time on the long-term one, that's OK with me.
|
lilmo
|
|
response 108 of 111:
|
Apr 13 02:43 UTC 1999 |
Re #82: Why would you drop foot bones off a shelf, and then try to break them
with a very old hard drive? :-)
|
scott
|
|
response 109 of 111:
|
Jun 6 17:47 UTC 1999 |
Guess this is the best item for this:
Great news! We have 2 donated SCSI disk enclosures which would be ideal,
except that they run on 240 volts which is not available in the Pumpkin.
However, it turns out that the power supplies in these enclosures are good
for as low as 100 volts, so that we should be able to use these just fine.
Right now we have 2 disks balanced on top of a cheap little box... not very
elegant.
(I fiound this out while rummaging around during a backup)
|
janc
|
|
response 110 of 111:
|
Jun 7 04:27 UTC 1999 |
Are these the big flat 6-drive enclosures? Cool. I think we have a
spare power supply for these too.
|
scott
|
|
response 111 of 111:
|
Jun 7 10:53 UTC 1999 |
Well, it looks like only one of them is flat and holds 6 drives, but yes.
|