Original question:
> System: Alpha 3000/300, OSF/1 3.2 214
>
> [output of "df -k /var/spool/news":]
>Disk usage:
>Filesystem 1024-blocks Used Avail Capacity Mounted on
>news_domain#news 8220432 3784112 1927416 66% /var/spool/news
>
> [output of "showfdmn -k news_domain":]
> Id Date Created LogPgs Domain Name
>328770e2.000d0bc0 Mon Nov 11 13:30:58 1996 512 news_domain
>
> Vol 1K-Blks Free % Used Cmode Rblks Wblks Vol Name
> 1 2055240 401344 80% on 128 128 /dev/rz11c
> 2L 2054976 256848 88% on 128 128 /dev/vol/newsvol1
> 3 2054976 515624 75% on 128 128 /dev/vol/newsvol2
> 4 2055240 753600 63% on 128 128 /dev/rz18c
> ---------- ---------- ------
> 8220432 1927416 77%
>
> So how much space is in use in the news spool -- 66%, as reported by >"df
-k", or 77%, as reported by "showfdmn -k"?
I got two answers:
1) Lars Bro <lbro_at_dscc.dk>, who surmised that I had more than one fileset
in news_domain, and that the df command as I'd given it was showing me only
the space taken by the "news" fileset at /var/spool/news while showfdmn was
showing me the space taken by all the filesets.
That was my fault for asking an inexact question: I explained that there's
only one fileset, "news", in news_domain, so the output of both commands
should be identical.
He then asked if there had ever been other filesets in the domain that may
not have been completedly deleted (there hadn't), warned that the
filesystem may be on its way to a crash (a fear that I share, if only
because it may think it's run out of space when it really has 2 gigs
somewhere it could be using) and suggested running a backup, zapping the
file domain, reinstalling it and restoring it from backup (which, given the
length of time this would take and how voluminously Usenet News flows,
would amount to just blowing it all off and starting from scratch -- which
I may yet have to do).
He also pointed me to an unsupported utility, /usr/field/vchkdir, that
checks the directory structure on an AdvFS volume. I ran it and it reported
no errors, but it didn't affect the disparity between df and showfdmn.
2) Jef Hamlin <jef_at_bhe.com>, who offered two solutions: "One is a patch
which should correct the problem and is available via download from DEC
(freebie). I believe the patch ID is OSF360-061. My call was worked by
Lannie Loh if that helps.
"The temporary workaround until you can get the patch is to enable quotas
(it's okay if you don't use them, turn them on anyway) and run vquotacheck.
This will bring the numbers closer and prevent your system from hanging (it
will eventually hang if one of these things is not done as it "thinks" it
has run out of space)."
So I tried the workaround first. I corrected /etc/fstab to read
news_domain#news /var/spool/news advfs rq,userquota,groupquota
throttled the server, turned quotas on with "vquotaon news_domain#news" and
ran "vquotacheck news_domain#news". It took five hours to complete and ...
df and showfdmn are still 10% apart:
# df -k /var/spool/news
Filesystem 1024-blocks Used Avail Capacity Mounted on
news_domain#news 8220432 3661708 2270448 62% /var/spool/news
# showfdmn -k news_domain
Id Date Created LogPgs Domain Name
328770e2.000d0bc0 Mon Nov 11 13:30:58 1996 512 news_domain
Vol 1K-Blks Free % Used Cmode Rblks Wblks Vol Name
1 2055240 406352 80% on 128 128 /dev/rz11c
2L 2054976 267936 87% on 128 128 /dev/vol/newsvol1
3 2054976 575600 72% on 128 128 /dev/vol/newsvol2
4 2055240 1020560 50% on 128 128 /dev/rz18c
---------- ---------- ------
8220432 2270448 72%
I'm now off in search of the patch.
========================================================================
Gordon T. Thompson gordy_at_nytimes.com
Manager, Internet Services 212 556 1386
The New York Times fax: 212 556 1636
This letter has been modified as follows from its original
version: It has been formatted to fit your screen.
Received on Fri Apr 04 1997 - 01:19:11 NZST