|
|
| Author |
Message |
| 25 new of 147 responses total. |
steve
|
|
response 75 of 147:
| 
Mar 28 15:14 UTC 1998 |
Well, the new machine has been up for 151 hours as I write this
(6 days 7 hours). Given that we've had an active week, I'm very
pleased with the stability we've seen so far.
It's likely that we're going let the machine run 'till something
happens to it. On the Sun-4/260 with SunOS 4.1.3 things were fine
'till about 11 - 13 days of uptime. We then decided that weekly
reboots were the best thing. On this new system, things might be
more stable. Jared was poking me at the last board meeting, reminding
me that I wasn't feeling charitble towards 4.1.4 before, which is at
least partially true. I'd heard verious comments about it being
unstable in vvarious situation, but that doesn't seem to be the case
with Grex's usage of it.
So we'll see what happens.
|
arthurp
|
|
response 76 of 147:
|
Mar 29 05:16 UTC 1998 |
The new Grex is a wonder! Better than the last big machine change.
One note on the software list above. make 3.76.1 is broken. It
makes a mess on really large builds like libc. 3.75 works. You might
want to keep an eye on this.
|
mdw
|
|
response 77 of 147:
|
Mar 29 05:55 UTC 1998 |
So far as I could tell, gnu make was just plain broken; "make -n"
has never worked properly for me with gnu make. I use bsd make instead.
|
steve
|
|
response 78 of 147:
|
Mar 29 23:45 UTC 1998 |
We had an interesting situation Saturday night, about 8:25PM.
Damon paged me, saying that the load average on Grex had risen to
228 (it took him three minutes to do an 'uptime' command). I
thought that we were udner some sort of fork bomb attack, but
it turned out to be because of a problem with someone sending
lots of mail to a particular place, which then bounced like
crazy and Grex was dealing with an impromptu sendmailfest which
drove the load average through the roof. The interesting part
is that it cleared up *very* quickly: in under 10 minutes the
load average went from 200+ down to 1, after someone cleared
the mail queue up. So now, we know that while the load average
can indeed surge to incredible heights, it can recover faster,
too.
So, if you were on any time Saturday night from about 8:20
to 8:30pm, thats why Grex seemed broken.
|
other
|
|
response 79 of 147:
|
Mar 30 22:32 UTC 1998 |
i have one thing and one thing only to say on the subject: WOW!
|
steve
|
|
response 80 of 147:
|
Mar 31 03:29 UTC 1998 |
It's certainly been interesting to see how this CPU is working
out. We've had 211 hours of uptime right now (8 days, 19 hours).
A little earlier today we had a blivit of CPU usage, sush that
the load av hit 17 for a brief period of time. Within 5 minutes
either way of that peak we were somewhere around 1. So we still
get hit every once in a while, but the system recovers far faster
than the poor little 4/260 did.
|
davel
|
|
response 81 of 147:
|
Mar 31 14:20 UTC 1998 |
Not definite enough, STeve. That should be "far far far far faster" ...
|
steve
|
|
response 82 of 147:
|
Mar 31 15:05 UTC 1998 |
Well, yeah.
|
remmers
|
|
response 83 of 147:
|
Apr 3 12:31 UTC 1998 |
I wish to announce that there are only two people logged in:
headdoc ttys2 Apr 3 07:16 (204.212.46.131)
remmers ttys9 Apr 3 07:24 (204.212.46.131)
Shades of 1992!
|
headdoc
|
|
response 84 of 147:
|
Apr 3 12:37 UTC 1998 |
We can't be the only two early birds, can we? My early wakening is not by
choice. I am at my office by 10 to seven as a condition of employment.
Getting on Grex makes it a wee bit easier to tolerate.
|
scott
|
|
response 85 of 147:
|
Apr 3 15:03 UTC 1998 |
The ISDN line is down, hence the only people logged on are locals. :(
|
rcurl
|
|
response 86 of 147:
|
Apr 3 16:18 UTC 1998 |
It's busier now...
biohazar 8r{> ta Apr 3 10:46
illogic Marjory K Hood *p7 Apr 3 11:08
mstone Mark Stone *rf 12 Apr 3 10:58
rcurl Rane Curl *ra Apr 3 11:13
The dialin lines were once very busy...what has happened that they are not
used as much? Has everyone just shifted to the "superhighway" and left the
old road to grow weeds, even though it is perfectly useable (and less
crowded)?
|
dpc
|
|
response 87 of 147:
|
Apr 3 16:48 UTC 1998 |
Hm. If there *is* a real reduction in dialin use, maybe we should
disconnect a few.
|
scott
|
|
response 88 of 147:
|
Apr 3 16:56 UTC 1998 |
Daily usage patterns... *all* the dialins get used at least once a day, if
only for a few minutes at the daily peak around midnight. During the workday
the dialins are much less used.
|
scott
|
|
response 89 of 147:
|
Apr 3 17:02 UTC 1998 |
Interesting... I just checked the groupie stats, and it looks like more people
are now using hardware flow control (which is correct) than last time I looked
(groupie keeps track of xon/xoff).
Unfortunately, groupie has to be rebooted to reset the port use counters,
which of means losing any connections that are active. This makes it hard
to track daily usage. I could print out each day's stats and do subractions,
I suppose. Right now I *am* sure that each line gets some use on a regular
basis.
|
keesan
|
|
response 90 of 147:
|
Apr 3 20:17 UTC 1998 |
I am now one of the people using hardware flow control, which Scott pointed
out that I needed in order to transfer binary files with X, Y, or Z modem.
(Actually I am not at the moment, but I do when transferring files - Jim put
the other commun software in a different directory, harder to get at.)
I have not been able to dial in about once a week when it is busy, but rarely
had to wait more than 5 minutes for a free line.
|
senna
|
|
response 91 of 147:
|
Apr 4 04:18 UTC 1998 |
Last night, at one point there were five poeple logged in, and three of them
were steves.
|
garima
|
|
response 92 of 147:
|
Apr 4 07:11 UTC 1998 |
Steve invasion
|
valerie
|
|
response 93 of 147:
|
Apr 4 14:15 UTC 1998 |
This response has been erased.
|
mdw
|
|
response 94 of 147:
|
Apr 5 00:52 UTC 1998 |
The fixwait binary last changed 9 january 1998.
|
valerie
|
|
response 95 of 147:
|
Apr 5 15:39 UTC 1998 |
This response has been erased.
|
mdw
|
|
response 96 of 147:
|
Apr 6 21:58 UTC 1998 |
Yup, fixwait would certainly switch between your first & second formats
depending if there are remote users or not. It should often also say
"; %d head" if there is someone in the wait queue. Also, fixwait will
attempt to use another language if you have the LANG variable set to a
valid locale. Valid locales include Af_ZA, De_DE, Es_ES, Fr_FR, It_IT,
and Pt_PT. Note however that csh has its own ideas of what a valid
locale is.
|
jerome
|
|
response 97 of 147:
|
Apr 7 17:40 UTC 1998 |
What is the purpose of the "head" value in the fixwait command's output?
|
srw
|
|
response 98 of 147:
|
Apr 7 18:06 UTC 1998 |
So that all of the processes (one for each one on the queue) can tell
when they are at the head of the queue. the process at the head of the
queue does a lot of different checking than processes that are waiting
to get there.
|
valerie
|
|
response 99 of 147:
|
Apr 9 14:11 UTC 1998 |
This response has been erased.
|
