At 11:01 AM 6/10/02 +0200, you wrote:
>Hiya!
>
>I've got some problems lately with dnews refusing to accept new posts due
>to diskspace problems. I got the following errormessage:
>
>10 10:04:34.017 0:error: lib: Not enough space, stopping feeds Used
>215641MB=100%, Free work 1083Mb, hist 1083Mb, spool 1083Mb, log 2773Mb
>** History area full (less than history.dat size free) 1172607 > 1083861
>
>Now my questions:
>
>Does "diskspace" include diskspace used by history and work data or just
>the plain postings (spool area)?
It depends, if they are on the same disk/partition, which they are in your
case, then yes.
> Can I reserve space for the history and the work directories by lowering
> the diskspace value?
As above, yes in your case.
>---
>
>History_min should always be the same as the current space occupied by the
>history file, shouldn't it? Is it possible to automate this setting e.g.
>dnews automatically pausing itself if it isn't able to write the history
>file after an expire? (i ran into this problem when i restarted dnews, it
>refused to accept new posts because the history area was full, i started
>an expire run, dnews accepted new posts via feeds while still expiring,
>after expire was finished the space didn't suffice for the "new" history file)
It is automated, however, during expire, the space is shrinking while the
history
file is being re-written, as a result it must allow less space during
expire than
it does when an expire is not running, until we did this perfectly stable
systems with
sufficient disk space to work normally would 'pause' the incoming feed
every night
for no reason.
Normally this doesn't create a problem, unless your expire process is
deleting less
data than you receive during one expire run, for this to be true you must
not be
expiring data very fast, so it will almost certainly block up very shortly
anyway.
A rebuild_index will unjam your system, or decreasing diskspace setting
which I suspect
you need to do anyway.
>What about workarea_min/spool_min?
Those areas do not have critical limits like the history area, anything
more than zero
free is workable.
>-
>
>What happens to the disk space used by "deleted" groups? After doing a
>"rmgroup <somebigbinarygroups>, purgegroups, exit, start, expire" i get a
>lot of messages like that in the expire.log:
>
>10 10:52:36.897 0:info: Couldn't find group {chile.binarios.misc}
>
>Do I have to rebuild the index after every major group-deletion?
No, you have to do a rebuild_piles after deleting groups if you want the
space to be released, if you don't care about the space being released you
don't
need to do a rebuild, the space will eventually be released as the buckets
rotate
normally. How long this takes depends on your system, it may be 2 days or
2 months.
WARNING, REBUILD_PILES IS BROKEN IN RECENT VERSIONS PRIOR TO 5.6 so DO NOT
do a rebuild_piles without first upgrading :-)
ChrisP.
