Doug:
1) This should have gone to [EMAIL PROTECTED] This response is
forwarded there.
Anyways..
> o Although the token said it was still valid, it appears to have
> expired a few hours early. This may be a time zone problem?
> The PC clock is set to the local time so I can boot DOS.
> I used in "rc" "clock -s" so as to get Linux time set correct.
> Are the AFS routines doing something different here?
> I need to look at this again, since I may be something else.
Linux-AFS uses the kernel system clock for current time. So long as
"date" returns the current time (and timezone) and "date -u" returns
GMT, then that should be working. Make sure that this was all set
_before_ you start Linux-AFS. If "date -u" does not give you GMT
properly, then this is the problem and you need to fix your system
clock to be correct.
> o "df" shows the cache never got above 10% full. I have a 100MB
> cache partition, and set the cacheinfo file to have 100000
> I would have expected all the source to endup in the cache.
I don't know -- I've seen a 10M cache get to be 95% full. Use "fs
getcache" to see how much of the cache is really in use. Maybe you
just aren't loading a lot into the cache? (I don't know, exactly).
Also, I do know of a bug in the stat cache; vcache entries are being
flushed too often. I don't have any more details at this time,
however, so I can't explain it much more that that without more
investigation.
> o To put a load on it and see how it works,
> I started building Kerberos 5.4.3 which is located in AFS. There
> where a number of lost contact with file server messages,
> which cause the make to get an error and continue on causing ,
> other problems. Over the weekend, I tried several times to get
> the make of the libkrb5.a to work, and it would always fail
> with some type of error while adding the object decks to the lib.
This is a documented problem in the Notes that come with the
distribution. If you want to help me track down this problem, I would
be most grateful (I realize that without sources it becomes more
difficult). I'm not sure if this is Linux-AFS directly or possibly
related to a bug in AFS 3.3...
> There may be ISDN problems as well, but are there any AFS or
> TCP/IP settings which could be set to extend the time before
> there was a lost contact message?
Not that I know of.. As I said, this is a known bug in Linux-AFS. I
don't know the cause, and I haven't been able to track it down, yet.
-derek
