In message <9509211015.AA28595@ncc.ripe.net>, Geert Jan de Groot writes:
Even 1G seems not to be enough. Any hints???? /db: write failed, file system is full dbm store returned -1, errno 28, key "3267231744/16" at /home/xladm/lib/cld b.
On Thu, 21 Sep 1995 00:01:31 +0200 (MET DST) Arnold Nipper wrote: pl line 134, <db> line 382118.
Hi Arnold,
So far, we are unable to reproduce this problem at the RIPE NCC. We have: $ ls -ls *in.cl* 96 -rw-r--r-- 1 auto-dbm 475136 Sep 20 18:18 ripe.db.in.cl.dir 6896 -rw-r--r-- 1 auto-dbm 1728316416 Sep 21 11:02 ripe.db.in.cl.pag
Note that while the 'size' of the file is large, it only takes 7MB of real disk space (the UNIX 'hole file' feature), not anything near what you are reporting.
$ df . Filesystem kbytes used avail capacity Mounted on dbase.ripe.net:/export/nccfs18/Dbase 759381 471792 211651 69% /tmp_mnt/ncc/dbase These are the log files since eternety, mirrors + indexes of the other databases, testdatabases and other gunk.
In order for us to be able to reproduce this problem, we would like people that encounter it send reports to ripe-dbm@ripe.net with the following information:
1. ls -ls of the files involved 2. Do you 'clean' the database frequently? 3. perl -v? 4. ls -ls of the bin directory of the database 5. egrep $Id of the same bin directory 6. You OS & version number, patches? 7. If you know, what dbm you are using with perl Help save the world & send us reports.
Hope this helps,
Geert Jan & David
This is almost certain to be due to the version of *DBM that you have linked in with perl. Try linking perl with the BSD DB library, using the NDBM emulation. Curtis