|
|
This item is for system problems. If something on Grex isn't working right (line noise on a modem, weird behavior from a program, etc.), this is the place to announce it. Except for security holes. If you find a hole in system security, mail information about it to "staff".
124 responses total.
As the staff probably knows, /var is full, /tmp is full. I am finding that I can't edit conference responses I'm writing.
Sounds like Grex is really full of it. 8-)
Is there some way to tell my total grex memory usage? (home directory, www, and mail directories)
The unix command is 'du'. This gets you the size of whatever directory tree you give it. You'll have to do this for your home directory, and for your mail file (which I am not sure that du can isolate from the rest; you might want 'ls -l /var/mail/keesan' for that. Memory use, I'm not sure of, but 'free' gets the total memory use of the system, including both real and virtual. There might be an option to get memory use only for processes owned by you.
To see the size of her mail file, Sindi should execute
!ls -l /var/spool/mail/k/e/keesan
I did the var/spool/ thing and got about 92K. I then typed du /a/k/e/keesan and got 226 for mail 16 for www 551 for home directory Does the var/spool/ thing show only mail in inbox?
Yup.
(du shows disk usage in 512-byte blocks. Multiply by 512 to get bytes.)
Or (on Grex, anyway) you can use du with the -b option to get bytes.
but be aware that the total for "du -b" and "du -k" doesn't necessarily
jibe..
% du -s -k ~mcnally
45 /a/m/c/mcnally
% du -s -b ~mcnally
29208 /a/m/c/mcnally
My guess is that "du -b" just adds up the bytes in the file contents,
which is not actually an accurate way to count how much space is being
used by the files.
Files in the 4.2BSD (McKusick, et al.) filesystem being used on Grex
consume filesystem space in greater-than-1-byte increments. The
filesystem is divided into "blocks", which are themselves sub-divided
into "fragments". Depending on how the filesystem was initialized,
the smallest amount of disk space a file can consume will be one fragment
(disregarding the file's entry in a directory file, which also consumes
some space.)
That means that if Grex's filesystems are created with 4K blocks, each
consisting of 8 fragments (I don't know what the actual parameters are,
I didn't create the filesystems when they were set up..) that the smallest
amount of disk space a file can use will be 512 (= 4096 / 8) bytes(*)
So if you have a file with just a single character in it, and assuming
I am right about "du -b", then "ls -l" will report a size of 1 byte,
"du -b" will count it as a byte, and "du -k" will count it as a full
fragment -- 512(*) bytes -- a much more accurate measure of how much
space you're really using.
(*) again, assuming 512 byte fragments *only* for purposes of example.
/a and /c were built with a fragsize of 1024 and a blocksize of 8192.
(out of curiousity, why is there no /b ?)
There is a /b. It's the bbs shell. The bbs shell was given the shortest possible name to save space in the password file.
/b is a symbolic link to /bbs/bbs.sh, which is the "real" bbs shell.
Right, that was created back long before we saw that we needed a more general way of labeling the partitions for user areas. The current one letter system will let us have 23 more such partitions, and hopefully we'll never go beyond that! So poor little /b is a lagacy directory.
"*lag*acy" meaning it increases system lag? 8-{)]
Munificent Grace to ye grex abacus staff whom in one day sothe to an conquered a dreaded Y2M monster inhabiting the sacred plane of the grex abacus.
This response has been erased.
Is there a reason I wouldn't receive a couple of e-mails that a particular person sent me? I believe in both cases he mailed me by responding to an e-mail I sent him. It happened over a week apart and I'm getting plenty of e-mail from other people.
I apparently sent someone half an email yesterday (subject line was the last thing that got sent) from a computer that we discovered shortly after had a virus. He could check for viruses. This is a partition table virus, whatever that means.
Don't worry. Partition table virii can't be spread by email, or at least I don't know of any that can be. And there's nowhere to stick one before the subject line anyhow.
partition table viruses?
I was suggesting that the friend who had trouble sending email might have a
problem at his end. (See the item on viruses for a detailed description of
our virus problem.) And that he should check his computer for viruses. A
virus may have been responsible for an intermittent formatting problem on the
computer that our virus spread from.
Partition table viruses can definitely spread by floppy disk and by
nullmodem cable. Anyone know about tape backup? If so, please offer us some
help in the computer virus item (66, I think it is). We would like to know
how to remove the virus from the hard disk, since repartitioning and
reformatting did not remove it.
(would someone please explain to me what good it would do to put a virus in the partition table? are you, perhaps, talking about *boot sector* viruses?)
sdscan called it a partition table virus. It also infects the boot sector of a floppy disk. It does spread by tape, too. Anti exe virus is what sdscan calls it.
re 24: On a hard disk the partition table is in the boot record. Or rather the first boot record. There is one per disk plus one per partition.
re #26: I still don't understand what good it'd do putting virus code
in the partition table. It's not like the partition table is
executed the way the boot loader is.. If, by "partition table
virus" people mean a virus that overwrites the partition table
with stuff (virus code, garbage, or anything else..) then I
guess I understand the point of that (though it's a nasty thing
to do..) but it doesn't sound like that's what Cindi's talking about.
It hides there and attacks exe com and dll files. It is gone now, Kent fixed it. Now Netscape often runs, but at 120 bytes per seco or less, as low as 30. What has hit grex since midnight? Very slow.
<sigh>
Re #27: I'd guess that'd be the case. Bear in mind most people have no clue about how their hard disks are *really* laid out.
Was there a grex problem from 12 to 12:20 or so this morning? If not, we have some other phone or modem or computer problem. Please let me know. Our two Netscape computers were running at as low as 68 bytes/sec and grex was waiting 10-15 seconds to display a word.
Around 12:06am, Grex gets VERY busy for a bit, doing a bunch of nightly things. There have been a few occaisons when I thought Grex was under attack, it was so slow. This is something we should work on optimizing; the work needs to be done, but hopefully we can alter the schedule such that there isn't an emmense queue of things to do all at the same time.
What happens and where at 12:06? India is 10 1/2 hours different.
I'm not sure how that relates to what Steve said ... at all ...
There is a program called cron which can run other programs at a specified time, in order to do things like simple system maintenance, and other things. A bunch of log files get either updated or moved every night, among other things. The way things have evolved, right at midnight a bunch of things fire off at once, which wasn't by intent but has come about that way, as we've added new things for cron to do. One example is that Grex now makes copies of the all-important passwd and shadow files six times a day in multiple locations, such that if the /etc directory is ever damaged we have a copy of the passwd database no more than 4 hours old. But all this takes a little time, and I think we're doing too much at once right at midnight. We'll have to look at all the things that are done, and work on rescheduling them a little bit such that we can spread them farther apart.
re 34, I thought all the Indian grexers might be reading email at once.
Re. 33: Good point. Maybe time should Grex should use GMT as the std. time reference.
GMT no longer exists. Try UCT.
Isn't GMT still the timezone Britain is in, even if it's not the official timezome of the world anymore?
| Last 40 Responses and Response Form. |
|
|
- Backtalk version 1.3.30 - Copyright 1996-2006, Jan Wolter and Steve Weiss